<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Miki Shapiro on Medium]]></title>
        <description><![CDATA[Stories by Miki Shapiro on Medium]]></description>
        <link>https://medium.com/@mikishapiro?source=rss-c6c055d9e96d------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*gqwXVYqdM-fraNVq.</url>
            <title>Stories by Miki Shapiro on Medium</title>
            <link>https://medium.com/@mikishapiro?source=rss-c6c055d9e96d------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 09:42:19 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@mikishapiro/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Steering: How Baking Decisions into Legal Contracts Helps Projects Fail]]></title>
            <link>https://medium.com/accountable-design/steering-how-baking-decisions-into-legal-contracts-helps-projects-fail-90d6ca94216?source=rss-c6c055d9e96d------2</link>
            <guid isPermaLink="false">https://medium.com/p/90d6ca94216</guid>
            <category><![CDATA[accountable-design]]></category>
            <dc:creator><![CDATA[Miki Shapiro]]></dc:creator>
            <pubDate>Tue, 02 Feb 2021 20:20:42 GMT</pubDate>
            <atom:updated>2021-02-05T02:47:53.346Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1023/0*0O2JakAvxTdkEvlL.jpg" /></figure><p>On December 7th, 2020, an article bearing news of the passing of one Chuck Yeager, possibly the most celebrated US Air Force test pilot of all time, caught my eye. Yeager, first man to break the sound barrier, came to be regarded as the cream of the crop in this highly dangerous and demanding occupation in the 1950s and 1960s, when jets were new and human safety an afterthought. As a kid, I grew up wanting to be an airman, and the ethos Yeager created had <em>everything</em> to do with it.</p><p>It seemed strange to many people when, shortly after the formation of NASA in 1958, and having opted to recruit their astronauts from the ranks of test pilots, NASA decided not to consider test pilots lacking tertiary education for the space program, and in doing so, eliminated some of the best pilots they had, including Yeager himself.</p><p>But there was another reason Yeager didn’t end up being one of the Mercury Seven (or an astronaut at all), as we learn from the thorough interviews Tom Wolfe had with him when researching his book “The Right Stuff”.</p><p>Yeager didn’t want to.</p><p>He based this off a more substantial reason than a bruised ego. Being canned and launched at mach 30, something even a lab animal could (and did) do, felt like a demotion from the absurdly demanding job of piloting untested aircraft at mach 1–3, something that even most trained pilots couldn’t.</p><p>Airplane developers, (eg “<strong><em>People Who Have Done This Before</em></strong>”) knew one thing really well. They knew that entering an untested set of conditions, there are things that are <strong>unknowable from a planning vantage point</strong>. Things that will make the wings fall off. Quite literally, in their case. The only thing they can do about it at the planning phase is to design in a maximally <strong>enabled operator</strong>, who would <strong>make decisions</strong> as the issues arise, and <strong>steer </strong>away from catastrophic outcomes. Yeager was already being entrusted to be this operator, in his existing test pilot job, on the leanest, meanest, fastest machines of his time.</p><p>Meanwhile, the designers of the original Mercury program capsule, NASA’s first jab at a manned spacecraft, had a different view. They worked on a loose assumption that most if not all things<strong> can be predicted and resolved in the planning phase </strong>(or perhaps, a fatalistic view that all failures are implicitly catastrophic). A <em>successful</em> execution of the plan was the be-all and end-all of their scope. The craft would be flown from the ground, with delays this causes being inconsequential as all instructions would be planned. To them, an astronaut was dead weight, an unneeded reality TV actor, placed in their craft for public relations.</p><p>Allow me to drive this point in. When the Mercury team went to inspect the capsule, they found out it had <strong>no manual controls</strong> designed into it <strong>at all</strong>. Or windows. Or even a way to exit if rescue didn’t show up. I swear, I am not making this up.</p><p>A mini-mutiny by the astronauts followed, and manual pitch/yaw/roll controls, a window and an escape hatch were added, averting what could have spelled a very different trajectory for NASA’s future. In Chuck Yeager’s recollection, though, NASA would remain in the business of putting a monkey in a can — though to NASA’s credit today, this righted itself and got better.</p><p>Fast forward to the present, to the world of IT services. We too now face this same choice.</p><p>We call an IT vendor, decide loosely what we intend to do, and all but etch those decisions into stone, by writing them into a <strong><em>legal contract</em></strong> as a <strong><em>Statement of Work</em></strong>.</p><p>Now, this is no moon landing. At best, we spend days, not years, figuring out what we will do in the event everything that might go wrong will go wrong. We just get going, Murphy’s Law be damned. And due to the heterogenous complexity our IT environments, we hit new challenges with near certainty, and they all lead to one core question: “Can we easily change early decisions to the best available real-world path leading to our goal?” … “<strong>Can we easily <em>steer</em>?”</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/479/0*i6GCIFZI-QWwOvAm.jpg" /><figcaption>Now What?</figcaption></figure><p>Now, the point I am trying to make here is not “AGILE good, waterfall bad”. Yes, iterative response to change is right in the AGILE manifesto. And all projects, even the waterfall ones, have a means by which to steer. For sure. The core questions I want to ask are:</p><ul><li>By whom is the steering done? The team on the ground? An executive who sits outside the team? A designer who is no longer sufficiently around?</li><li>How much <strong><em>total</em></strong> effort does a steering action require to run its course?</li><li>Given the required lead time (time elapsed from identification to resolution) and process time (people’s time consumed) a steering action requires, <strong><em>how many times</em></strong> can a steering action realistically occur over the course of our project?</li><li><strong>If we could make this number higher, would we?</strong></li></ul><p>“Mr Yeager… here’s the test aircraft. We’ve designed it so over the course of your flight, you can make a steering adjustment no more than 4 times. Use them wisely”. Sounds absurd, doesn’t it?</p><p>So how do we do this better?</p><p>When more than one commercial party is involved, <strong>steering </strong>lies squarely in the lap of the legal contract we write. It will determine if steering remains closely guarded by “the people designing the craft” (plan/design team), or “the people piloting it” (delivery team).</p><p>On the topic of steering, our legal contracts loosely divide into two flavours:</p><p>Those of <strong><em>flavour A</em></strong> that say “<strong>You <em>delivery</em> team people are <em>not</em> trusted. We the <em>planners</em> know better. You shouldn’t be steering without an adult in the room. We will weigh down any steering you need to do with change control, sign-offs by time-poor people not fully familiar with nuanced context, adding extra cost and extra negotiation effort for every attempt to steer; We will try to make steering happen LESS</strong>”. It usually reads something like “The team will deliver a solution and the solution will <em>look </em><strong><em>like this</em></strong>”.</p><p>Then there are those of <strong><em>flavour</em> B</strong> that say “<strong>You delivery team <em>are</em> trusted, we recognise your vantage point gives you a better view than us planners, and you understand issues arising we cannot see yet. You CAN steer, you SHOULD steer, and we, the contract writers invested in your success, will remove all hindrance. We will try to make steering happen MORE and contractually set up the best conditions for it.” </strong>Such conditions are often set up in the form of AGILE ceremonies, such as sprint planning sessions, demos and retros. Statements of work of this flavour may read <em>“The customer will deliver six 2-week sprints towards goal X, whereby the customer Product Owner will attend sprint planning sessions and carry ongoing responsibility grooming the backlog feeding the sprints”</em>. Though the goal is specified, solution shape is deliberately not mentioned — to make any required steering need to jump through fewer hoops, and thus happening more.</p><p>Those two approaches solve two very different problems.</p><p>Flavour A, the Mercury v1.0 Flavour, solves for commercial liability in case a vendor catastrophically fails to deliver on a project. A rare event (particularly with large experienced vendors, where long term relationships and industry reputations are in play). And it solves it not by fixing anything, but by issuing the team weak steering tools and a very big threat.</p><p>Flavour B , The Yeager Flavour, solves for a much more common real world problem— unexpected hurdles. It places clear ongoing responsibilities not just on the vendor but the customer too (who typically, at a minimum, appoints a Product Owner to manage priorities as an ongoing effort), and provides far better correction and early exit pathways than the nuclear “We will sue you for failing to meet your commercial obligations” option. It expects the delivery team to steer.</p><p>Ironically, Flavour B solves for catastrophic failure better, by introducing early interventions to right the ship rather than the threat of a nuclear option after too much money has already gone down the drain, and things are far past correcting. Additional agreed steering mechanisms often seen in flavour B allow rotating vendor resources when a customer is dissatisfied with their skills, for example, or simply an option to terminate the engagement earlier (eg “stop buying sprints”) if the customer wishes to abort early.</p><p>A common argument for Flavour A can be boiled down to “hope for the best but plan for the worst”. The problem with that argument is that you’re not just <em>planning</em> for failure. By taking steering away or making it harder, you are <strong><em>explicitly creating</em> the conditions that make failure more likely</strong>. Flavour B does the opposite. By making steering desired, easy and baked into routine work, it commercially mandates our teams are well set up to respond to issues.</p><p>There is no us and them here, these two choices don’t favor a customer over a vendor or vice versa. As customers <strong>and</strong> as vendors — would we rather set the project up, to be able to steer its way to success?</p><p>Or would we rather, as Chuck Yeager put it, put a monkey in a can, make the project harder to steer, and have slightly bigger nukes for the blame game that follows a failed project?</p><p>—</p><p><em>“Test Pilot Charles Elwood ‘Chuck’ Yeager in a Lockheed NF-104 Starfighter, Dec. 4, 1963” by aeroman3 is marked under CC0 1.0. To view the terms, visit </em><a href="https://creativecommons.org/licenses/cc0/1.0/"><em>https://creativecommons.org/licenses/cc0/1.0/</em></a></p><p><em>“tree perception 1” by Foxtongue is licensed with CC BY-NC-SA 2.0. To view a copy of this license, visit </em><a href="https://creativecommons.org/licenses/by-nc-sa/2.0/"><em>https://creativecommons.org/licenses/by-nc-sa/2.0/</em></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=90d6ca94216" width="1" height="1" alt=""><hr><p><a href="https://medium.com/accountable-design/steering-how-baking-decisions-into-legal-contracts-helps-projects-fail-90d6ca94216">Steering: How Baking Decisions into Legal Contracts Helps Projects Fail</a> was originally published in <a href="https://medium.com/accountable-design">Accountable Design</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Manage a Self Managed Team]]></title>
            <link>https://medium.com/stuff-that-needs-many-humans/how-to-manage-a-self-managed-team-cf52914e46cc?source=rss-c6c055d9e96d------2</link>
            <guid isPermaLink="false">https://medium.com/p/cf52914e46cc</guid>
            <category><![CDATA[stnmh]]></category>
            <dc:creator><![CDATA[Miki Shapiro]]></dc:creator>
            <pubDate>Wed, 16 Dec 2020 00:54:58 GMT</pubDate>
            <atom:updated>2020-12-18T04:47:04.180Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/996/1*1u3RKU_kGhYwXcYASwq1tQ.png" /></figure><p>Let’s talk about what it takes to get a self-managed team going.</p><p>Suppose you have staff, and want to solve a messy problem that hasn’t been assigned to anyone. And you don’t just want to solve it well, you want your shop to be a rewarding place to work and grow. But all while solving your messy problem.</p><p>Do it with a <em>self managed team.</em></p><p>A bit tricky to set up, but… as teams go… the most effective kind there is.</p><p>Here’s how it works —</p><ul><li>You need to start with <strong>an aim</strong>.</li><li>We will assume you have <strong>authority</strong> to form such a team (though in some organisational models, this can be kicked off even without authority).</li><li>You need a group of people who have demonstrated they <strong>deeply care</strong> about the aim. Yes, this means you can only proceed if your aim is something people care deeply about.</li><li>At least one (but really as many of them as possible) needs to have <strong>an idea of what it looks like when it’s done right</strong>.</li><li>At least one (but really as many of them as possible) needs to have a high <strong>emotional intelligence</strong>.</li><li>It really helps if at least one has experience in <strong>skills, techniques and practices that help team formation</strong> and difficult conversations, skills like how to write a social contract together, or resolve a disagreement.</li></ul><p>You put them in a room.</p><p>You tell them you want their help with the aim.</p><p>You tell them you want to hear <strong>what it would take</strong> to reach that aim, <strong>what they’ll need</strong> and <strong>what they come up with</strong>. But not <em>right then</em>.</p><p>And then, since you represent authority, <strong>you have to walk out</strong>.</p><p>Oh, and <strong>do not put anyone in charge of that group</strong>.</p><p>Dan Pink, author of the book ‘Drive: The Surprising Truth About What Motivates Us’, popularised the idea that what motivates people most is not just money (<a href="https://www.youtube.com/watch?v=rrkrvAUbU9Y">money works weirdly</a>), but the trinity of <strong>purpose</strong>, <strong>autonomy</strong> and <strong>mastery of their art</strong>.</p><p>What we just did by putting our cocktail of people together was <strong>selected for mastery and purpose</strong> — we picked people with knowledge who care about our aim, and we <strong>created autonomy</strong> by letting them work out how they intend to approach the problem.</p><p>Now the fun starts. What we just put in a room is humans, not robots. Let’s be realistic about what to expect of a group of common, passionate and smart humans.</p><p>Some time back, Bruce Tuckman, a psychologist studying group dynamics, determined that teams go through four stages of team growth — <strong>Forming</strong>, <strong>Storming</strong>, <strong>Norming</strong> and <strong>Performing</strong>.</p><p><em>Forming</em> is where the team gets to know each other, or finds out they’ll be working together towards a goal.</p><p>Then comes, <em>Storming. </em>We cherry-picked people who care about our aim. People who care get emotionally invested, and often, attached. When everyone on that team doesn’t know where everyone else stands, things can get messy, involve high passions, misfires, hurt egos, even people getting strong urges to leave.</p><p>It is in these moments, where it helps having members with high emotional intelligence, empathy and leadership skills to help the team overcome internal chafing and form a common identity and a common goal.</p><p>Three important things to remember:</p><ol><li>Disagreement, heated discussions and personality clashes <strong>are</strong> <strong>common</strong>.</li><li>So long as the team manages it, it’s <strong>not bad</strong>. It’s a formative stage.</li><li>So long as the team manages it, it’s <strong>temporary</strong>. It’s a formative stage.</li></ol><p>So long as the relationships in the team strengthen as a result and people find ways to work together, this helps form resolve.</p><p>The third, <em>Norming</em> stage is when we start making progress but without the friction. The team members start to better understand each other, find ways to accept each other’s limitations and develop a sense of safety around each other’s motives and modus operandi.</p><p>The fourth and last, <em>Performing</em> stage happens when the team not only knows how to not step on each other, they start actively leveraging each other’s passions, strengths, expertise, resulting in real progress towards the overall aim.</p><p>For this reason — once you have such a team, keeping them in that zone is critical. Excessively shuffling people will simply knock a team back to earlier formation stages, resulting in less time spent in high gear getting stuff done.</p><p>Finally, touch base with the team with one simple question: <strong>What can I do to help?</strong></p><p>One such team I‘ve recently been a member of was tasked with setting up an event for some 150 people (none of us were professional event coordinators).</p><p>Our leadership team, e.g. those who set this team up, did several things that we appreciated:</p><ol><li>They were not intrusive. They asked when it would make sense for them to be involved, and asked to be guided by the team.</li><li>They allowed the team to adjust (read: broaden) the aim where possible.</li><li>They offered help, <em>but did not force it</em>. Help with resources, networking, experience and advice.</li><li>They let the team make its own decisions.</li></ol><p>This autonomy came back as high performance work, like a boomerang with a vengeance.</p><p>There is something very human-compatible and incredibly gratifying about working with people who care deeply about the same thing. And not just sharing a passion, but sharing a purpose, a goal, and working together to reach it.</p><p>We humans do the best work under these conditions.</p><p><em>— <br>Thank you to Marlou Rees and Dr Julia Richardson for peer review and editorial help with this article.</em></p><p><em>&quot;Storming over Two Trees” by …-Wink-… is licensed with CC BY-NC-ND 2.0. To view a copy of this license, visit </em><a href="https://creativecommons.org/licenses/by-nc-nd/2.0/"><em>https://creativecommons.org/licenses/by-nc-nd/2.0/</em></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cf52914e46cc" width="1" height="1" alt=""><hr><p><a href="https://medium.com/stuff-that-needs-many-humans/how-to-manage-a-self-managed-team-cf52914e46cc">How to Manage a Self Managed Team</a> was originally published in <a href="https://medium.com/stuff-that-needs-many-humans">Stuff That Needs Many Humans</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[From Chess to Open Source: Can Our Stories Become Universal?]]></title>
            <link>https://medium.com/stuff-that-needs-many-humans/from-chess-to-open-source-can-our-stories-become-universal-5328a252bdc3?source=rss-c6c055d9e96d------2</link>
            <guid isPermaLink="false">https://medium.com/p/5328a252bdc3</guid>
            <category><![CDATA[stnmh]]></category>
            <dc:creator><![CDATA[Miki Shapiro]]></dc:creator>
            <pubDate>Sun, 06 Dec 2020 03:55:04 GMT</pubDate>
            <atom:updated>2020-12-16T00:59:19.063Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/768/1*eskzK60OuLca4boi1UPXLw.jpeg" /></figure><p>What connects Yuval Noah Harari’s book ‘Sapiens’, ‘The Queen’s Gambit’ show on Netflix and Open Source?</p><p>Perhaps, but only perhaps, it is the elusive idea of <strong>a universal in-group story</strong>. A story about us, that <em>has no out-group</em>. One that has <em>no need</em> for a ‘them’ to define or promote ‘us’.</p><p>How many such stories do you know? Virtually all religions have an out-group. So does every national story. So does any class-based society.</p><p>A universal in-group story is a bit of a holy grail, as big aspiring stories go. And yes, they’re in extremely short supply. The book “Sapiens” <em>is</em> one such story. And although chess culture and Open Source are not, and may never become such a story, for reasons ranging from access to interest to identity, each overcame an obstacle that brought them a significant step closer. Mid-20th century chess culture overcame an identity hurdle, while open source pulled down the access bar.</p><p>Whether these two ever become “universal in-group” stories we all identify with isn’t the point. Probably not. But it’s what we can find in them, an ingredient, a pattern we may wish to look for in the other stories too, that this article is about.</p><p>Yuval Noah Harari’s book — Sapiens, tables what is possibly the biggest “universal in-group” story of our day. I use it here to set a high water mark. In it, Harari shows what it means to be human. Us sapiens — and no other species — use fictions we believe at the same time, as a tool to cooperate in large numbers. We’ve been doing that for some 70,000 years. We do it flexibly. Unlike bees, whose cooperation is hardcoded in DNA, we adjust our stories and respond to change. In the Q&amp;A sessions following Harari’s talks, an interesting comment gets made again and again — people from all over stand up and just tell him “the story you told here — I feel it’s mine”. Our stories — of nations, corporations, religions, money, even road rules — enable millions of <em>strangers</em> to work together. I won’t go any deeper into these ideas, but rather, explain why people from <a href="https://www.youtube.com/watch?v=AnPs8vnZ0I4">President Barack Obama</a> to <a href="https://www.gatesnotes.com/Books/Sapiens-A-Brief-History-of-Humankind">Bill Gates</a> vocally promote this book. The reason? The universally relatable, novel, evidence-driven story about how <em>we</em> actually roll. He did it not by creating some new religion, but by simply pointing out what any bundle of ideas — religions not exempt — really are (in a good, meaningful way), and made everyone scratch their head, look at their church or national flag, soccer club or car brand, favourite ancient civilisation or Harry Potter books, and go “Shit. <strong>He’s right</strong>. That <strong><em>is</em></strong> us”.</p><p>Rather than elevate Sapiens onto some ‘big history’ pedestal, reserved for grand narratives and global ideologies, we’ll look at it differently — a story that achieved a black belt with something to teach, facing numerous other stories that aspire to something (perhaps not that grand) themselves, and may wish to learn.</p><p>Let’s jump to ‘The Queen’s Gambit’. I have a personal connection to concede. I come from a Russian home. My mum reached “Pervi Rozryad”. Quarter finals. <em>National</em>. In what was then Leningrad, USSR, 1975. Because reasons — health complications and emigration, that would be the end of her bout with competitive chess. I am not competitive, but I saw many a chess board growing up, and watching the Netflix series brought it home for me. This does translate into some sense of identity, and that positionality does bias my views.</p><p>“The Queen’s Gambit” is a fictional story set in the chess culture of the 1950s and 1960s. Hidden in that chess culture lies a story not only with <em>universal</em> <em>access</em> aspirations — we will talk more about access when we look at open source — but an <em>identity story</em> too. A story I wholeheartedly feel is mine as well. An identity that seems to be about a board game, but really isn’t. It’s really about what one can do armed with the human mind, not about pieces of wood.</p><p>Let’s take a minute to acknowledge the faults of that culture that made it less universal. To its credit, The Queen’s Gambit does not avoid them. The elitism in the sport was probably worse in real life. The show peppers it in, though it smoothes things off, offering an entire buffet of ways to gain admiration, by showing true grace on the harsh stage of defeat. You know, should anyone in 2020 happen to need one.</p><p>Elitism wasn’t the only major barrier to universality. Sexism was, too. While Beth Harmon, the heroine of The Queen’s Gambit does her victory lap in that white queen suffragette dress, no such heroine actually existed — the real person around whose struggles Beth was modelled <a href="https://www.goodhousekeeping.com/life/entertainment/a34645948/queens-gambit-true-story-beth-harmon/">was Bobby Fischer</a>, who, according to all accounts, was not a woman. The show makes for good “time-machine catharsis”, in that very “Inglorious Basterds” way. We send an agent of present thinking into the unjust past, and watch them fix history and win. In ‘Inglorious Basterds’, Quentin Tarantino sent a squad of ruthless Jews to make the Nazi war machine regret its life choices. Similarly, The Queen’s Gambit sends a 9 year old budding female chess grandmaster into the 50s to overcome misogyny, the prejudice of low expectations women got from the closest people in their lives, ultimately to face off the Russians at the thing they are notoriously best at. Cathartic. But fiction. Reality wasn’t like that for women, not in the US, and by my mum’s account, not in the USSR either. In the last episode, The Russian female champion drives it home, when it’s revealed she doesn’t play the men. Chess, back then at least, left women behind.</p><p>But look closely, and you notice something that did actually happen. Chess culture is a small identity story — a game, not an ideology. Yet it penetrated through not one, but two divisive identity barriers, the iron curtain, and an extreme rich/poor class divide. How?</p><p>“Hang on”, I hear you say. Didn’t, like, all sports do that? The Olympics? Soccer?</p><p>Yes, they did. And no, not like that they didn’t. The difference? The fruits of physical training can‘t be shared. Knowledge can. Two types of actors in the chess community — those running competitions, and the masters winning games and writing books — despite all the competitive elitism in the world — <em>did systematically share</em> <em>knowledge</em>. And it was shared not among friendly team-mates. It was shared <em>across cold war borders.</em> Between the most bitter of enemies. Using government money. It was shared with <em>everyone </em>who had access to a medium-sized library. An outline of an <em>us</em> starts to appear…</p><p>What was it the chess community was sharing with this in-group? What did it mean to <em>be us</em>? And is there an <em>us</em> here, that doesn’t need a <em>them</em>?</p><p>Unlike many other sports, chess action doesn’t all happen in the game itself. The competitive chess players don’t play, win and party. They play, meticulously recording their moves. They don’t just want the outcome at the end. Like an algebra teacher, they seek the method, not just the answer. They want a way to go back and change it. They want <em>the source</em>. The Queen’s Gambit does this justice. Time and time and time again, they go back to games they played, and play them again with variations. Study each other’s past games. They learn, from teachers or literature, known sequences and strategies — openings, mid-games and end-game plays. They <em>analyse the open recipes</em> precisely describing past games, the sharpest looking for a place to diverge, improve on their predecessor, for a better outcome. And when they succeed, reaching fame and glory, their accomplishment is put back, sharing those recipes. Sharing them with <em>us</em>.</p><p>Chess culture cracked an identity problem. It did so through a system that provided its broad base of non-competitive players a universally shared purpose — the mastery of a tool we all have — the human brain. A public good provided by the competitive sphere — recipes for games, was knowledge any one of us could use to hone our own mind. I want to identify with such a group. Many of us do.</p><p>If you come from the open source world, you’ve already spotted the shared DNA that chess world has with your own. It dovetails into our third story. <em>Our</em> third story. Who is <em>we </em>this time?</p><p>Anyone who wants to create something with technology. Anything. Anywhere. Whether you’re the CEO of Microsoft or a bright 19 year old entrepreneur in Bangalore, the implied, free toolbox you start out with says Open Source on the cover. What you’ll find in it are not the world’s poorest tools. They are the world’s finest, never mind those developing most rapidly.</p><p>You, and I. What can we do? Just the two of us? For absolutely any combination of us?</p><p>We can prototype a billion dollar solution with a bill of materials that has 3 figures. We can install an enterprise-grade operating system on virtually any computer. Run an enterprise-grade web server. Write a computer program, and compile it in any number of computer languages. Run distributed computing technology like Kubernetes or Hadoop — technologies that undergird Google’s infrastructure. Automate work that used to need an army to do. And that’s not where you and I are going. That’s just where <em>we get to start. </em>Today that is.<em> </em>Pop the lid open a year from now, there’ll be even more.</p><p>I said open source solved an access problem. That wasn’t strictly true. Like with chess, where we had to see the full system — normal people using words like <em>we</em> and <em>us</em>, while competitive elite openly share knowledge — here too, we need to look at the big picture, and acknowledge what role open source plays in the wider system.</p><p>In recent decades, not one, but many things beat down the access barrier to technology everywhere. x86 compatible hardware. Standards. Moore’s Law. One Laptop Per Child and similar projects, putting tech in the hands of millions of kids for a price developing nation governments can handle. Skills acquisition on Youtube. Massive open online courses. First Sourceforge, then Github and friends. Kickstartr. Raspberry Pi and other ARM computers. Microlending. The list goes on and on. Are we there yet? Have <em>we</em> finally solved the problem of <em>them</em> who simply don’t have access? Not by a long shot. And the complexity of what we are trying to give everyone access to just keeps increasing, moving the goalposts further and further afield.</p><p>Did you notice I didn’t mention open source in that list? Not because it doesn’t have a role, but because its role is unique. It didn’t just give everyone with youtube tuition and a cheap device access to highly complex software. It gave us <strong>something to do with it all</strong> once the bar got low enough. Open source helped establish <strong><em>demand</em></strong><em> </em>for all those other things, by making the One Laptop Per Child projects, the Raspberries, the affordable phones mean not “a device” but “all the things open source software can already do”. At scale. “<a href="https://www.youtube.com/watch?v=-PDZv3f20Dg">Open unlocks the world’s potential</a>” we say where I work. That’s <strong>literal</strong>. Open source closed the circuit, sending the entire system into a positive feedback loop in overdrive. That’s a very big part of how we got to where we are today.</p><p>Let’s come back to universal in-group stories. What does it mean for <em>us</em>? If we drag that access bar way down, where every new thing done by those who now have access, stands not at the foot of some immense commercial competitor, but on the shoulders of our collective software achievement, at scale, for every starter upper in the world?</p><p>All our societies push forward entrepreneurs, movers, disruptors, challenging them with our nastiest problems. <em>We all</em> have some big problems we want solved, right? Our soldiers are only as effective as the tools we send them into battle with. Well, so are our entrepreneurs. A lower access bar means more starters, more battles won, oh, and a bigger collective pile. I signed up.</p><p>I wish we had more universal in-group stories. They are not just emotionally empowering, they help — directly, by helping achieve a purpose, and indirectly, by doing away with out-groups, and in doing so reducing antagonism, conflict, friction and resistance. They don’t have to be perfectly universal, even part way goes a long way, so long as they’re headed towards, rather than away from, universality.</p><p>Can you think of more?</p><p><em>— <br>Thank you to Dr Julia Richardson for peer review and editorial help with this article.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5328a252bdc3" width="1" height="1" alt=""><hr><p><a href="https://medium.com/stuff-that-needs-many-humans/from-chess-to-open-source-can-our-stories-become-universal-5328a252bdc3">From Chess to Open Source: Can Our Stories Become Universal?</a> was originally published in <a href="https://medium.com/stuff-that-needs-many-humans">Stuff That Needs Many Humans</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Four Vendor Designs. A Goldilocks Story.]]></title>
            <link>https://medium.com/accountable-design/four-vendor-designs-a-goldilocks-story-fcca2ea0ae8e?source=rss-c6c055d9e96d------2</link>
            <guid isPermaLink="false">https://medium.com/p/fcca2ea0ae8e</guid>
            <category><![CDATA[accountable-design]]></category>
            <dc:creator><![CDATA[Miki Shapiro]]></dc:creator>
            <pubDate>Thu, 12 Nov 2020 22:47:16 GMT</pubDate>
            <atom:updated>2020-11-14T22:13:13.520Z</atom:updated>
            <content:encoded><![CDATA[<p>Hello.</p><p>My name is Fipo, and I am a fictional product owner.</p><p>Today is my last day in this role, and I want to tell you about my legacy here.</p><p>Over the past 30 months, I have been the product owner of several technology solutions, working in the IT department of a large bank. I have several ops teams report to me. I myself report to an executive, who in turn reports to the CIO.</p><p>Over my tenure, I engaged four technology vendors, to help my team solve four big problems.</p><p>I don’t control the world. Quite on the contrary, to get anything done, I had to find allies, buy-in and agreement among other groups in the bank, ranging from my internal providers, to my internal customers, to oversight teams like cybersecurity, and even my own teams, as well as align my efforts with the goals set by my boss and my CIO.</p><p>I want to tell you about the four High Level Designs, or HLDs, left by the design phases from my engagement with each of the four vendors.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qDjzw71jC7RHPmKPadOPpg.png" /></figure><p><strong>The first one</strong> was the rockiest.</p><p>Our vendor wrote a design, and it wasn’t until the design was complete that we realised nobody agrees or even wants it. The vendor did not make an effort to seek out all the stakeholders and listen to them (some of this is my own fault, but I’ll be the first to admit, we simply didn’t know how to ask the right questions), or tease out our requirements. Not just the business requirements I provided, but the full, nuanced set of solution-shaping forces that would need to be respected and addressed for it to work.</p><p>We had to throw that HLD in the bin and start again. Twice.</p><p>I honestly felt like I needed to pay consulting rates to teach the vendor how to do their job. It was here I first understood that writing a design without agreement is like putting wrapping paper around nothing at all.</p><p>It was not a pleasant experience, and I would not care to repeat it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Iju5rhz9MvJkO-UOSLHevA.png" /></figure><p><strong>The second one</strong> was better, but, honestly, not by much.</p><p>The solution was there, and so was the agreement on requirements and decisions. It was passable. Except… I regretted asking the vendor to write the document the moment I saw it. It was plastered, head to toe, with the vendor’s <em>stuff</em>. Their other products, their irrelevant features. It was someone else’s marketing brochure. One I allowed to be written on our letterhead on my watch, signed in my name, and focused far more clearly on a third party’s interests than on mine.</p><p>I still remember how I sat there, embarrassed, with our architecture review board, to discuss their feedback on the design, so as to get it approved.</p><p>It was not making me or my team look good, let me tell you. It gave every impression we’re in an imbalanced relationship with the vendor, and not only was trust in this vendor sinking as a result, my team was going down with them. The architects knew it, and I knew they knew.</p><p>That HLD got approved. After it was built, I put it away, intending never to look at it again, or have to relive the embarrassment.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*E4JxGdkn1xBkmNEQPc27Uw.png" /></figure><p><strong>The third one</strong> was different again.</p><p>Here too, we formed agreement about requirements and decisions. And it didn’t radiate evidence of my team being beholden to the vendor either — though it gave us a bad run in a different way.</p><p>You could spot the problem immediately if you looked at it printed out. Sideways.</p><p>It was a dense, dense beast. Much larger than it needed to be, and it didn’t tell the story of the solution, so much as aiming to be a testament to how smart the author was. And while the vendor was evidently proud of the work they’ve done, the energy and commitment needed of anyone I engaged to look at my solution was prohibitive. People just gave me the “too hard” glaze-over.</p><p>It took my architecture approval process a full two months to digest it, compared to the typically 2–4 week process we usually see. Other teams didn’t care to look at it. It was just too dense. One of the architects gently told me that he doesn’t need that level of detail in an HLD, and maybe I split it out to a separate detailed design document. My engineer told me they felt some of the detail could have been worked out during implementation, and that preemptively detailing it all was overkill. Another colleague mentioned some of it was just pasted in product documentation, which, they felt, would have better been left online as the information dates, and was simply used as “padding”. I think I made a face when I heard that.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OVtPH-7r1TGbwfkh46pnsQ.png" /></figure><p><strong>It was the fourth and last one </strong>that restored my faith in there being such a thing as doing it right. A Goldilocks design document.</p><p>It was <strong>rooted in agreement</strong>, with all the people whose resistance or non-participation could conceivably slow it down. And it was like this <strong>because the vendor would not have it any other way</strong>.</p><p>It <strong>captured the requirements</strong> and <strong>cited them in its decisions</strong>. Each time we had to make a decision and no requirement guided it, the vendor would pull us up. “If we can’t justify the decision with a requirement… we forgot to list the requirement.”</p><p><strong>It was not a large document.</strong> It had a clear essence. A purpose. A clear scope. A problem statement. Decisions, each naming what other options were considered. Anyone who picks it up in two years can instantly assess if those decisions are still relevant to them then.</p><p>The architect approval was in and out of this one in a week. Both because it was concise, and <strong>because we engaged them early</strong>. In fact, some of the requirements there were their own words. <strong>This was their baby too.</strong></p><p>Not all was smooth. Reality intervened and we thought we’d have to build it sub-optimally. But the vendor understood the consequences, and explained them, calling out assumptions and dependencies in the design itself, helping word the risks register and reflect it back to the business. The help understanding and expressing those risks got the business onboard and paved a path forward. It’s all in the design.</p><p>Most critically, this design — it was not about the vendor, or their story. <strong>It was my story</strong>, and how our team overcame a big challenge. It felt good that more than anyone, the vendor knew this. They understood the focus was on our own journey. And they had the clarity at each step.</p><p>It’s a bit of a guilty pleasure, but after showing it to several teams, one asked me to introduce them to this vendor, as they had a related problem, and they felt they couldn’t handle it as well as we did on their own. I caught myself showing off a bit. And as I looked around — it wasn’t solely my baby. Every one of the stakeholders treated it like their success too.</p><p>And that is exactly the kind of thing I got into leadership for in the first place.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fcca2ea0ae8e" width="1" height="1" alt=""><hr><p><a href="https://medium.com/accountable-design/four-vendor-designs-a-goldilocks-story-fcca2ea0ae8e">Four Vendor Designs. A Goldilocks Story.</a> was originally published in <a href="https://medium.com/accountable-design">Accountable Design</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How Much Does Agreement At Work Cost Us?]]></title>
            <link>https://medium.com/accountable-design/how-much-does-agreement-at-work-cost-us-3ca02253582a?source=rss-c6c055d9e96d------2</link>
            <guid isPermaLink="false">https://medium.com/p/3ca02253582a</guid>
            <dc:creator><![CDATA[Miki Shapiro]]></dc:creator>
            <pubDate>Fri, 23 Oct 2020 01:38:28 GMT</pubDate>
            <atom:updated>2020-10-26T21:09:16.544Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/241/0*8c_ca81ZtAXjcq-K" /></figure><p>It’s an important question to ask. Not because we shouldn’t pay the price, but rather, more often than not, precisely because we should.</p><p>Paying people to forge agreement carries a cost. Forgoing it carries a far higher one.</p><p>My own context is helping drive organisational change where I work, in a large tech firm, and what I argue here applies whether you’re driving it from the bottom, the top, or from anywhere in between.</p><p>More broadly it applies to any plan that involves more than one stakeholder.</p><h3>Setting Out to Make a Change</h3><p>Suppose <strong>Sally</strong>, <strong>Ben </strong>and <strong>Dan</strong>, managers of three operations teams, want to make a big change.</p><p><strong>Nicole</strong>, their boss, controls their budget — which is currently insufficient to make the change.</p><p>In the past, <strong>Ben </strong>strongly argued for it, but other priorities tabled by <strong>Michael</strong>, who manages sales, won out.</p><p>This budget cycle, Sally, Ben and Dan decided to play it differently.</p><p>They first formed an agreement among themselves on how to proceed. It was no longer “Ben’s” case to make to the boss. It was all three of them. They understood that speaking on behalf of their groups with a unified voice gave them more sway.</p><p>They also well understood Michael’s team won’t stop being a business priority. Hell, they want Michael’s team to succeed.</p><p>So they figured out a way their change can benefit sales. They sat down with Michael’s group, argued their idea, changed it as necessary, and worked it until Michael’s team got behind it.</p><p>Then they all, Michael in tow, showed up in Nicole’s office.</p><p>Usually, that’s when stuff changes.</p><h3>What Did They Pay for Agreement?</h3><p>I want to draw attention to two core ideas.</p><p>FIRST POINT: What they did — forging informed, discussed agreement — is <strong>work</strong>. This work carried a cost — time they committed.</p><p>Reaching agreement — like any process — carries a lead time and a process time.</p><p>SECOND POINT: Looking at agreement from the outside — as a <strong>commoditised task,</strong> that <strong>takes time</strong>, <strong>has a cost</strong>, and <strong>whose prime benefit is removing an obstacle </strong>or <strong>seeking buy-in</strong>— we can start reaping some systemic benefits:</p><ol><li>We can plan better — with safer evidence-based assumptions (evidence being written agreement), neither the lack of agreement nor the cost of reaching an agreement take us by surprise.</li><li>We can methodically set up the right structures to support reaching agreement.</li></ol><h3>Narrowing Down the Definition</h3><p>Let’s set some definition boundaries. The ‘agreement’ discussed in this writeup implies:</p><ol><li>All parties being informed on the matters they see as relevant.</li><li>All parties <strong>feel safe</strong> to dissent.</li><li>All parties had an opportunity to think things through. To borrow vocabulary from Daniel Kahneman’s book ‘Thinking, Fast and Slow’ (here’s <a href="https://www.youtube.com/watch?v=uqXVAo7dVRU">a great primer</a>), everyone had the opportunity to get into ‘System 2’ thinking.<br>Humans employ two thinking systems, a rapid low-cost one that takes many shortcuts (System 1) and an expensive one that can handle complexity and long-term objectives (System 2). The agreement we discuss here needs the latter.</li><li>All parties consenting (often to compromise). Consent at gunpoint is not agreement. But consent through recognising another party’s perspective and willingness to compromise from a place of shared goals is.</li></ol><h3>Guiding Principles to Healthy Agreement</h3><p>UNIVERSAL IN-GROUP</p><p>From a problem-solving standpoint, it’s a ‘<strong>universal in-group</strong>’ story. There is no “them”. It’s all “us”. It doesn’t matter who reports to whom, or who controls the purse strings. We’re all operating this beast together, each from their seat, and we all deeply care about some aspect of it. Agreement is simply an increment of collective progress, no matter whose seat you look at it from.</p><p>ADDITIVE GAME</p><p>It’s not zero-sum.</p><p>Sally, Ben and Dan needed to go through two rounds of effort — a first round, to <strong>add up</strong> their individual ideas of success into something that works for all three of them. That wasn’t a given.</p><p>But then, they didn’t need Michael to lose funding to succeed. On the contrary, they needed him to gain something, to arrive with a broader coalition of allies. His objective was <strong>added</strong> to theirs.</p><p>AN EVENTUAL SOLUTION IS ASSUMED, and ITS SHAPE DOESN’T MATTER</p><p>Whoa. What?</p><p>A very human tendency we’re all guilty of is starting out with a specific solution in mind. This is, of course, a closed mindset (to equally good if not better options)… not to mention the subtle, seeping distrust we communicate to our colleagues when we don’t give them input into steering it.</p><p>The solution is the part that needs to be flexible. From an oversight point of view, we don’t need to dictate it, just ensure that everyone who needs to weigh in is present, and safe to do so.</p><p>We need to trust our people.</p><p>LEARNING NA’VI</p><p>I recently re-watched James Cameron’s movie ‘Avatar’ with my daughter, and a particular gem in it deeply resonated with me.</p><p>It was the Na’vi people greeting — “I see you”.</p><p>Norm, one of the human characters, gives the audience a casual 10 second deep dive, explaining it is always meant beyond the literal.</p><p><em>I can see your point of view.</em></p><p><em>That matters to me.</em></p><p>It frames everything that is to follow around mutuality and forging a shared perspective.</p><p>This underlying mindset, expressed in our real world words and deeds, is the most fertile soil from which agreement can grow.</p><h3>Summing Up — How We Can Use Agreement</h3><p>Agreement is desired. Its absence spells blocked work.</p><p>Agreement can be</p><ul><li>Commoditised</li><li>Systematically facilitated</li><li>Described as a process — that has a cost — and used to plan.</li></ul><h3>So How Do We Do This Right?</h3><p>Look at your plan. Whatever plan.</p><p>Ask:</p><ul><li>Did we factor in the agreements needed to eliminate all possible blockers? Or did agreement get assumed and its time and cost externalised?</li><li>Which agreements will we need to have the best shot at success?</li><li>With whom?</li></ul><p>And ask the systemic question too: Are we set up, with the required stakeholders and the necessary safe environment, to reach these agreements?</p><p>This makes or breaks almost everything we do.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3ca02253582a" width="1" height="1" alt=""><hr><p><a href="https://medium.com/accountable-design/how-much-does-agreement-at-work-cost-us-3ca02253582a">How Much Does Agreement At Work Cost Us?</a> was originally published in <a href="https://medium.com/accountable-design">Accountable Design</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Compiling a Custom Kernel on Fedora on Raspberry Pi]]></title>
            <link>https://medium.com/ironhaul/compiling-a-custom-kernel-on-fedora-on-raspberry-pi-9e199731220b?source=rss-c6c055d9e96d------2</link>
            <guid isPermaLink="false">https://medium.com/p/9e199731220b</guid>
            <category><![CDATA[kernel]]></category>
            <category><![CDATA[fedora]]></category>
            <category><![CDATA[raspberry-pi-4]]></category>
            <dc:creator><![CDATA[Miki Shapiro]]></dc:creator>
            <pubDate>Sat, 25 Apr 2020 23:17:46 GMT</pubDate>
            <atom:updated>2020-12-20T21:24:03.000Z</atom:updated>
            <content:encoded><![CDATA[<p>In the <a href="https://medium.com/ironhaul/installing-64-bit-fedora-on-the-raspberry-pi-4-d4a665ea65d3">previous piece</a> I showed how a current 64-bit Fedora distribution can be made to run on a Raspberry Pi 4. There was a catch though — we used the boot partition and the kernel from the Raspbian OS.</p><p>In this writeup we will replace this with our own kernel, and optionally add additional kernel features.</p><p>It is common to use an “initial ramdisk” — using either the <em>initrd</em> or <em>initramfs </em>scheme — an archive file containing a small filesystem and in it some drivers and tools — to equip a kernel at boot-time with extra capabilities (such as loading RAID or SAN drivers) before mounting the primary operating system disk. Adding this is possible, but is outside the scope of this article.</p><p>We will <em>not</em> use a mainline kernel from kernel.org, as some of the Raspberry hardware support is still only available in the <a href="https://github.com/raspberrypi/linux/">downstream repository</a> maintained by the Raspberry Pi people. We will instead use <a href="https://github.com/raspberrypi/linux/releases">their latest release</a>.</p><p>We will compile this kernel locally right on our arm64 Fedora, running on the Raspberry Pi 4 build in the previous writeup. This is an order of magnitude faster than using an arm64 emulator… and tidier than using a cross-compiler on a PC.</p><p>To keep this really simple, we will do this as root. And we’ll go old school and use <em>menuconfig</em>, because it’s just as awesome today as it was 25 years ago.</p><p>So, first things first:<br>Step 1: <strong>Grab the sources</strong></p><p>Browse <a href="https://github.com/raspberrypi/linux/releases">https://github.com/raspberrypi/linux/releases</a>, drill into the latest release (raspberrypi-kernel_1.20200212–1 as of the writing of this piece), and at the bottom of its release page, grab the link of the tar.gz package. Then, on your raspberry:</p><pre># cd /usr/src<br># wget <a href="https://github.com/raspberrypi/linux/archive/raspberrypi-kernel_1.20200212-1.tar.gz">https://github.com/raspberrypi/linux/archive/raspberrypi-kernel_1.20200212-1.tar.gz</a><br># tar zxvf <a href="https://github.com/raspberrypi/linux/archive/raspberrypi-kernel_1.20200212-1.tar.gz">raspberrypi-kernel_1.20200212-1.tar.gz</a><br>...<br># ln -nfs linux-raspberrypi-kernel_1.20200212-1 linux<br># cd /usr/src/linux</pre><p>Step 2: <strong>Install some tools</strong></p><pre># dnf -y install gcc flex make bison openssl-devel elfutils-libelf-devel ncurses-devel</pre><p>Step 3: <strong>Prepare a configuration file.</strong></p><p>If you followed the process from the first writeup, you should have Raspbian kernel’s driver modules under /lib/modules/4.19.97-v8+</p><p>We will load a nice little module from there called <strong>configs</strong></p><pre># modprobe configs</pre><p>As a result, /proc/config.gz will be available. This is compressed text containing your running kernel configuration (e.g. which bits to compile in, which to compile as modules on the side, and which to leave out). We will extract it straight into our kernel source tree.</p><pre>zcat /proc/config.gz &gt; /usr/src/linux/.config</pre><p>Run the <em>menuconfig</em> make target:</p><pre># cd /usr/src/linux<br># make menuconfig</pre><p>You can just go ahead and save the configuration as is. If it’s your first time doing this, I highly recommend you browse around. This gives an idea as to how expansive the Linux kernel is — including the many bits that don’t end up built into popular distributions like Fedora or Debian, and how much diverse stuff one can enable and get out of it, from drivers to packet filter capabilities, exotic filesystems, and upcoming technologies.</p><p>There are indeed many things you can do here — remove things you know you will never use — or add things you think you need. After you save your configuration, you will be ready to build.</p><p>For my nefarious purposes, I came here to add a few specific things — SELinux and CGroups. However, for these options to appear, their <a href="https://cateee.net/lkddb/web-lkddb/SECURITY_SELINUX.html">prerequisites</a> also need to be ticked in.</p><p>To add these particular ones via menuconfig, I went to <strong>Security Options</strong>, and enabled “<strong>Enable different security models</strong>”, and in there enable “<strong>Socket and Networking Security Hooks</strong>. In “<strong>General Setup</strong>” I enabled “<strong>Auditing Support</strong>”. Under “Security” again, I enabled “<strong>NSA SELinux Support</strong>”.</p><p>Of course, we can just add the lines you want to .config directly— namely, in the case of my stuff, this:</p><pre>CONFIG_AUDIT=y<br>CONFIG_AUDIT_WATCH=y<br>CONFIG_AUDIT_TREE=y<br>CONFIG_AUDITSYSCALL=y<br>CONFIG_AUDIT_GENERIC=y<br>CONFIG_AUDIT_COMPAT_GENERIC=y<br>CONFIG_INTEGRITY_AUDIT=y<br>CONFIG_INTEGRITY=y<br>CONFIG_LSM_MMAP_MIN_ADDR=32768<br>CONFIG_NETWORK_SECMARK=y<br>CONFIG_SECURITY_NETWORK=y<br>CONFIG_SECURITY_SELINUX=y<br>CONFIG_SECURITY_SELINUX_AVC_STATS=y<br>CONFIG_SECURITY_SELINUX_CHECKREQPROT_VALUE=0<br>CONFIG_SECURITY_SELINUX_DEVELOP=y<br>CONFIG_SECURITY_SELINUX_HOOKS=y</pre><p>I got the above list by running a diff between the original .config and the one I saved in menuconfig right after ticking in the SELinux stuff.</p><p>Of course, you can do a million other things in here, adding drivers, changing things like the XFS filesystem driver from being a loadable module to being compiled into the kernel itself, or removing dead weight to make your kernel ever so lean.</p><p>Once done, save your .config file.</p><p>Step 4: <strong>Build the kernel and its modules</strong></p><p>All we have to do now is build the kernel image (the uncompressed kernel <em>Image </em>make target — to keep in line with how the Raspberry people did it) and the modules (the <em>modules</em> and <em>modules_install</em> make targets).</p><pre># make Image <br># make modules <br># make modules_install</pre><p>Note that the Pi 4 has four CPU cores, so running several jobs in parallel can speed things up four-fold (optimum jobs being core count +1, so 5). I got what appears as a race condition on my first time doing this, (An <em>cp: cannot stat ‘./modules.order’: No such file or directory </em>error)… maybe because I ran multiple targets at once. This didn’t happen from the second run onward and sped things up substantially:</p><pre># make -j5 Image <br># make -j5 modules <br># make -j5 modules_install</pre><p>Another issue with this particular release is this error:</p><pre>[root@localhost linux]# make Image modules modules_install<br>...<br>modules_install/usr/bin/ld: scripts/dtc/dtc-parser.tab.o:(.bss+0x10): multiple definition of `yylloc&#39;; scripts/dtc/dtc-lexer.lex.o:(.bss+0x38): first defined here</pre><p>That’s really not a biggie. If you encounter it, fix it so:</p><pre>sed -i &#39;s/^YYLTYPE yylloc;$/extern YYLTYPE yylloc;/&#39;  /usr/src/linux/scripts/dtc/dtc-lexer.l</pre><p>… and re-run the make command to keep going.</p><p>This will run for a while and produce a new uncompressed kernel in /usr/src/linux/arch/arm64/boot/<strong>Image</strong> as well as modules in /lib/modules/&lt;kernel-name&gt;, which, in the case of the kernel used for this writeup, is in /lib/modules/<strong>4.19.97-v9</strong>. Note that since the original raspbian kernel calls itself “4.19.97-v9+”, its modules live in /lib/modules/<strong>4.19.97-v9+ </strong>next door to where we’re installing ours.</p><p>Step 5: <strong>Install the kernel</strong></p><p>The common way to “Install” this kernel — e.g. ask the active boot loader to offer it as a boot option at next boot — (make <em>install — </em>which runs the arch/arm64/boot/install.sh script) —<strong> we will not use here.</strong></p><p>This is because the install script doesn’t know how to handle the Raspbian vfat boot partition (which doesn’t support symlinks) and stuff will break if you run this. Keep in mind neither grub nor uboot is being used here as a boot loader. We will take the low-tech approach instead — we will simply copy the old kernel to a backup, and the new one on top of the old one’s filename.</p><pre># cp -vrfp /boot/kernel8.img /boot/kernel8.img.bak<br># cp -vrfp /usr/src/linux/arch/arm64/boot/Image /boot/kernel8.img</pre><p>We’ll also copy our new kernel’s System.map file to /boot to enable resolving kernel symbol names and addresses:</p><pre># cp -vrfp /usr/src/linux/System.map /boot/System.map</pre><p>That’s it.</p><p>Reboot the pi and you should be done.</p><p>To check it all works, make sure your pi boots and you can get back on.</p><p>Once you do, run</p><pre># uname -r </pre><p>To make sure the version of kernel you booted is the right one (in our case this will say “4.19.97-v9” if we succeeded, and bear the original Raspbian kernel name — “4.19.97-v9+” — if we got something wrong.</p><p>Then run</p><pre># lsmod</pre><p>If you get a list of modules that are already loaded, you’re golden. If you get an empty list, the kernel can’t load modules.</p><p>Congrats. Within the limitations of the Raspberry Pi kernel source tree, you are now in control of your kernel.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9e199731220b" width="1" height="1" alt=""><hr><p><a href="https://medium.com/ironhaul/compiling-a-custom-kernel-on-fedora-on-raspberry-pi-9e199731220b">Compiling a Custom Kernel on Fedora on Raspberry Pi</a> was originally published in <a href="https://medium.com/ironhaul">Ironhaul</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Installing 64-bit Fedora on the Raspberry Pi 4]]></title>
            <link>https://medium.com/ironhaul/installing-64-bit-fedora-on-the-raspberry-pi-4-d4a665ea65d3?source=rss-c6c055d9e96d------2</link>
            <guid isPermaLink="false">https://medium.com/p/d4a665ea65d3</guid>
            <category><![CDATA[raspberry-pi]]></category>
            <category><![CDATA[aarch64]]></category>
            <category><![CDATA[fedora]]></category>
            <dc:creator><![CDATA[Miki Shapiro]]></dc:creator>
            <pubDate>Sun, 05 Apr 2020 10:45:21 GMT</pubDate>
            <atom:updated>2021-03-10T04:29:31.774Z</atom:updated>
            <content:encoded><![CDATA[<p>Hacker; noun. One who makes furniture with an ax.</p><p>This is a hack.</p><p>This is a hack because as of the writing of this piece, we have two excellent halves of a solution, and I just need something that works.</p><p>On one hand, the Raspberry Pi Foundation have</p><ul><li>A pretty remarkable 1.5GHz quad-core, 4GB memory, 64 bit computer with enough stuff on the motherboard to qualify for a desktop, in a $50 package, called the Raspberry Pi 4b.</li><li>A Linux kernel that supports Pi 4b features.</li><li>A Raspbian (Debian derivate OS) image, which contains a boot loader that can controls the many hardware features of the Raspberry, just like a PC BIOS does. It also contains ready 32-bit and 64-bit Linux kernels, that will happily boot on any Linux distribution.</li><li>Sadly, they don’t maintain a 64-bit “userspace”, because they’re small and don’t want to support two userspaces, one for their 32-bit stuff and one for their 64-bit capable stuff.</li></ul><p>On the other hand, we have the Fedora project. An enterprise-grade operating system user-space built natively for 64 bit arm, with a massive catalogue of enterprise software, some Seriously Cool Stuff that can be built off it (I’m looking you right in the eye, Kubernetes+etcd straight from the OS repos). Sadly, Fedora does not currently support the Pi 4b, its u-boot boot loader doesn’t fit the Pi 4b like a glove, and the kernels as of this moment do not run stable on the Pi 4b model or support some of its hardware.</p><p>This guide will show how to put the above together in minimally working order. Because I need to repeat this a lot, I cut some Ansible code to do it all.</p><p>Some prep work first: <br>We will need one or more running PCs to fulfil the following roles (I assume you are running Fedora or a derived OS; it will work on Debian/Ubuntu just as easily, but you’ll need your nerd license to adapt it a bit):</p><ol><li><strong>base</strong>: A place to run some commands (an Ansible playbook).</li><li><strong>builder1</strong>: A place to assemble some images. All up, prep at least 16GB of space or so. This will be done as a user that can sudo to root.</li><li><strong>flasher1</strong>: A place you can insert an SD card. This too will need a user that can sudo to root.</li></ol><p>The same machine can be all three (but you can also use AWX/Tower for <strong>base</strong>, a virtual machine for <strong>builder1</strong> and a Fedora laptop for <strong>flasher1</strong>, if you’re thus inclined. It doesn’t really matter.)</p><ol><li>Install <strong>ansible</strong> and <strong>git</strong>:</li></ol><pre>[mikishapiro@base ~]$ <strong>sudo dnf -y install ansible git</strong></pre><p>2. Grab the code:</p><pre>[mikishapiro@base ~]$ <strong>git clone </strong><a href="https://github.com/mikishapiro/armbuild.git"><strong>https://github.com/mikishapiro/armbuild.git</strong></a></pre><p>3. Modify your default Ansible inventory file (<strong>armbuild/inventory</strong>), with the settings you want.</p><p>This includes naming which machine(s) will fill the <strong>builder1</strong> and <strong>flasher1</strong> roles, some configuration for the resulting Raspberry Pi Fedora image and so forth.</p><p>You will also need to plug your SD card in the <strong>flasher1</strong> machine, run</p><pre>[mikishapiro@flasher1 ~]$ <strong>sudo lsblk</strong></pre><p>… and figure out from the output what device it is. It’ll probably be a /dev/mmcblk* device if it’s sitting in an onboard SD card reader, or a /dev/sd* device if it’s sitting in a USB adapter. I use a USB adapter, and it appears as /dev/sdb on my system.</p><p>4. We’ll need to set up a bit of ssh access to allow Ansible to do the heavy lifting for us.</p><p>Ensure your user is in the ‘wheel’ group and can sudo to root on builder1/flasher1. To simplify this writeup, I will assume you can sudo to root without needing to provide a password (uncomment the %wheel line containing NOPASSWD in /etc/sudoers and comment the original one to achieve this).</p><p>5. Ensure your user on <strong>base</strong> and <strong>builder1</strong> has an ssh key generated. Run ‘<strong>ssh-keygen</strong>’ command and create one with a blank password if not.</p><p>6. Ensure your user on <strong>base</strong> can ssh to <strong>builder</strong> and to <strong>flasher</strong> by logging on to <strong>base</strong> and typing</p><pre>[mikishapiro@base ~]$ <strong>ssh-copy-id builder1</strong><br>[mikishapiro@base ~]$ <strong>ssh-copy-id flasher1</strong></pre><p>Also, you will need to ensure your user on <strong>builder1</strong> can ssh to root@<strong>flasher1</strong>.</p><pre>builder$ <strong>ssh-copy-id root@flasher1</strong></pre><p>That should be it.</p><p><em>Note: If you use any of this in any kind of production environment, you will need to make this method more secure, by storing sensitive configuration items (like password hashes or a public ssh key) in Ansible Vault (or AWX/Tower), or by avoiding the ssh to root operations I use. This is outside the scope of this writeup.</em></p><p>That’s the prep. Now just run the deploy.yml playbook:</p><pre>[mikishapiro@base ~]$ <strong>cd armbuild</strong></pre><pre>[mikishapiro@base armbuild]$ <strong>ansible-playbook -i inventory deploy.yml</strong></pre><p>You can add -vvv to the second command if you like gory debug.</p><p><strong>If everything is in order, that should be it. </strong>Some 10–15 minutes after the downloads are complete, you should have a flashed SD card (assuming USB2 transfer speeds). Transfer your SD card to your Raspie and power it up.</p><p>Re-running the playbook (with same or different SD card) will not re-perform the lengthy download operations if not necessary, and if you use the lazy=true configuration item in the inventory, it’ll recycle the existing unarchived images as well, saving more time.</p><p>Go ahead and have a look at the Ansible code to see what it does — the core logic is in the deploy.yml playbook, while some of the low level operations are inside the arm-build role, in the tasks/*.yml files related to the instructions it knows how to handle.</p><p>I’ve tested this using a pre-release build of Fedora Server raw-xz (aarch64) version 32 from here: <a href="https://www.happyassassin.net/nightlies.html">https://www.happyassassin.net/nightlies.html</a> and Raspbian Lite from here: <a href="https://www.raspberrypi.org/downloads/raspbian/">https://www.raspberrypi.org/downloads/raspbian</a></p><p>In a nutshell, the playbook (running on <strong>base</strong>) will</p><ol><li>Download and unarchive both OS images (to <strong>builder1</strong>), but without deleting the archives. There’s a bit of acrobatics in there for the Raspbian image to resolve its redirect URL, find out the real filename and save the image with the correct filename. This saves re-downloads.</li><li>On <strong>builder1</strong>, mount via loopback and modify the unarchived images <em>inside the unarchived files themselves</em>. In the case of aarch64 Fedora there is LVM on the image (this is different from 32-bit arm Fedora 32, which does not), and we need to install and use <strong>kpartx </strong>on <strong>builder1</strong> to allow lvm to see the contents so you can mount the filesystem on the logical volume inside. This will happen automatically.<br><em>Note: If the </em><strong><em>builder1</em></strong><em> system’s own LVM volume group is also named fedora (some, but not all, fedora versions do this), this playbook won’t be able to reference the contents of the loop device just by mentioning volume group name. You’ll need to do that bit manually, rename some volume groups, or fix some playbooks.</em></li><li>On the Raspbian image, in the boot partition (which contains all of <strong>bootloader</strong>, <strong>Pi 4b firmware</strong> <strong>configuration</strong> and <strong>kernels</strong>) we will tweak the firmware config (<strong>config.txt</strong>) to use 64 bit mode (this makes it boot using the kernel named kernel8.img in the same directory), and the kernel command line (<strong>cmdline.txt</strong>) to remove the once-off Raspbian repartitioning init directive (making the kernel default to /sbin/init).</li><li>On the Fedora image, we’ll change PermitRootPassword to yes in /etc/ssh/sshd_config, inject an ssh key to /root/.ssh/authorized_keys and/or a root password hash into /etc/shadow if specified in the inventory. Since we’ll be using the Raspbian kernel, we also need to bring from the Raspbian root partition the drivers that were compiled with it and for it. We will copy over /lib/modules/* from the Raspbian root partition to the fedora root partition, and bring along some modprobe configuration as well to prevent the wrong drivers from being loaded. We will also configure a network interface on the Fedora root (which appears in aarch64 Fedora as eth0, hence /etc/sysconfig/network-scripts/ifcfg-eth0) with IP settings you can set in the inventory. It can also disable some services to make the whole thing a bit quicker to come up.</li><li>Once the large image files on <strong>builder1</strong> contain what needs to go on the SD card, <strong>base</strong> will log on to <strong>flasher1 </strong>and do a bit of advance housekeeping, such as wipe the first 300MB of the SD card using /dev/zero to avoid any previous contents on the SD card from tripping up the use of the parted tool.</li><li><strong>base </strong>will now log on to <strong>builder1</strong> and using a ‘dd | ssh root@flasher1 dd’ command, stream the part of the Raspbian partition making up the Master Boot Record (which contains the Raspbian image partition table) as well as the boot partition to the SD card.</li><li><strong>base</strong> will now log on to <strong>flasher1</strong>, and remove the second (root) partition from the partition table (the data of that second partition isn’t actually on the SD card, we only streamed the MBR and the data for the first partition). It will then create a new partition on all the remaining space on the SD card, which, for the default Raspbian kernel to function, <strong>has to be ext4 and cannot be xfs</strong>. A new filesystem is then created on it, and it is mounted. <strong>base </strong>then proceeds to stream all the files from the Fedora image partition, still mounted via loopback on <strong>builder1, </strong>to <strong>flasher1</strong> using a “tar cfp — . | ssh root@flasher1 ‘cd &lt;sd&gt;; tar xvfp -’ ”command.</li><li>Some last rites — a chmod 755 on the root directory of the SD root partition on <strong>flasher1</strong>, tidy umount of the file images on <strong>builder1</strong>, kpartx &amp; losetup loopback device cleanup on <strong>builder1</strong> — and we’re all done. You’re set to re-run it.</li></ol><p>I hope you found this helpful.</p><p>In my next write-ups, I will be covering</p><ul><li>How to natively compile a kernel directly on the aarch64 Fedora v32 system produced above (instead of cross-compiling or using an emulator), using the kernel source tree provided by the Raspberry Pi foundation (which contains full support for the Pi 4b features), and with the fancy stuff we will need for OKD and COSA.</li><li>How to build the Fedora CoreOS Assembler (COSA) on this system, so we can assemble an aarch64 Fedora CoreOS image.</li><li>How to assemble an actual aarch64 Fedora CoreOS image.</li><li>How to build OpenShift Kubernetes Distribution (OKD) version 4.4 kube go components on our aarch64 system.</li></ul><p>Have fun, and don’t forget to be awesome :)</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d4a665ea65d3" width="1" height="1" alt=""><hr><p><a href="https://medium.com/ironhaul/installing-64-bit-fedora-on-the-raspberry-pi-4-d4a665ea65d3">Installing 64-bit Fedora on the Raspberry Pi 4</a> was originally published in <a href="https://medium.com/ironhaul">Ironhaul</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Red Hat Satellite and Ansible Tower]]></title>
            <link>https://medium.com/accountable-design/red-hat-satellite-and-ansible-tower-53c2cd1626f6?source=rss-c6c055d9e96d------2</link>
            <guid isPermaLink="false">https://medium.com/p/53c2cd1626f6</guid>
            <category><![CDATA[ansible]]></category>
            <category><![CDATA[soe]]></category>
            <category><![CDATA[automation]]></category>
            <category><![CDATA[satellite]]></category>
            <category><![CDATA[provisioning]]></category>
            <dc:creator><![CDATA[Miki Shapiro]]></dc:creator>
            <pubDate>Sat, 25 Aug 2018 00:28:02 GMT</pubDate>
            <atom:updated>2018-08-29T04:25:00.035Z</atom:updated>
            <content:encoded><![CDATA[<p>An integrating story.<br>by <a href="/service/mailto:mshapiro@redhat.com">Miki Shapiro</a> and <a href="/service/mailto:talor.holloway@gmail.com">Talor Holloway</a></p><p>Much has been written about <a href="https://www.redhat.com/en/technologies/management/satellite">Red Hat Satellite 6</a>, Red Hat’s long-standing enterprise server fleet management product, and <a href="https://www.ansible.com/products/tower">Ansible Tower</a>, Red Hat’s choice automation engine.</p><p>Despite obvious overlap between what both technologies do (configuration management, remote execution, node classification etc.), neither is a drop-in replacement for the other, and each is capable in ways the other is not.</p><h3>Why is this needed?</h3><p>Setting up infrastructure (be that physical, virtual or cloud) while providing teams with control planes accounting for <strong><em>everything</em></strong> that changes on each client system -</p><ul><li>the state systems are born into (provisioning)</li><li>the content, methods and cadence used to update them over time (patching)</li><li>what configuration and customisation they receive to be able do their job (CM)</li></ul><p>and</p><ul><li>what security bar they empirically meet (compliance)</li><li>… all while meeting the changing and evolving needs of their application workload</li></ul><p>… can be a daunting and complex task. By meeting these requirements, not only do we hit the lot, but to a degree, we divorce the scale from management costs by making <strong>all of it</strong> fair game for automation.</p><p>This article will discuss an integrated Red Hat Satellite and Ansible Tower solution architecture that addresses the following 4 (and optional fifth) requirements:</p><p>1. Use a <strong>single product to act as source of truth</strong> for facts driving provisioning, patching and configuration management.</p><p>2. Define all settings using a <strong>fleet hierarchy that supports nesting and inheritance</strong>.</p><p>3. Provide the user with a <strong>push-driven</strong> configuration management system.</p><p>4. Provide the user with workflow management that can <strong>perform actions on multiple systems</strong> within a single workflow.</p><p>5. (Optional) Often a direct preference requirement requests Ansible. If explored, the underlying requirement is either lower time/labor costs in deploying new or servicing existing automation, or intention to extend the solution to deploy immutable infrastructure for cloud-native applications.</p><p>Putting on our solution hats, Red Hat Satellite 6 hits the first two requirements head-on, however its native configuration management system, Puppet, fails on requirements 3–5.</p><p>Ansible Tower hits the last three, but trails behind when it comes to constructing all the content and server-side stuff a healthy client system needs, specifically for provisioning and patching.</p><p>Last, Satellite and Ansible Tower without integration fails requirement number one.</p><p>The solution meeting the lot is smart integration of the two products, leveraging the strengths of each.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*PirGvz7_vQ93rVrv" /></figure><p>For our purposes, the primary roles of Satellite are:</p><p>- Store and allow hierarchical management of configuration parameters/variables.</p><p>- Manage server side content (yum repos etc.)</p><p>- Provide low-level tooling for provisioning (e.g. PXE backend where appropriate, various types of boot images, etc).</p><p>whereas the primary roles of Ansible Tower are:</p><p>- Know everything Satellite knows about each system (using a Dynamic Inventory).</p><p>- Securely store and employ the credentials used to interact with external systems.</p><p>- Provide a configuration management system using the parameter/variable values from Satellite.</p><p>- Orchestrate user-defined workflows. For example, a workflow can enhance the Satellite’s provisioning capability by performing actions on other external systems before or after the actual deployment (create a VM, update a CMDB, update DNS, configure firewall rules etc.), or provision several different systems together as part of deploying a related product stack (for example, an OpenShift cluster).</p><h3>A High Level Overview of the Deployment Process</h3><h3>1. Deploy Satellite</h3><p>Red Hat Satellite should be configured with the appropriate Products, Content Views, Life Cycle Environments and Activation Keys.</p><p>Hostgroups are important in this process, as they define a host’s Life Cycle Environment, Content View, Content Source, Operating system specifics and parameters specific to the host that will be exposed to Ansible Tower as inventory facts. Hostgroups can be nested, allowing certain settings to be inherited from the parent Hostgroup by a child Hostgroup.</p><p>The Hostgroup tree should thus also be set up. For example, a single location/content-source tree for one operating system (omitting application customisation “leaves”) may look as follows:</p><p><em>soe-rhel7</em></p><p><em>soe-rhel7\sandbox</em></p><p><em>soe-rhel7\nonprod</em></p><p><em>soe-rhel7\prod</em></p><p>Finally, a service account should be created in Red Hat Satellite 6 to allow Ansible Tower to interrogate Satellite.</p><h3>2. Deploy Ansible Tower</h3><p>Ansible Tower should be configured to use Satellite 6 as a dynamic inventory source. Support for this is already included in recent versions of Ansible Tower, and can be set up by</p><ol><li><strong>Cog</strong> -&gt; <strong>Credentials</strong> -&gt; <strong>Add</strong> -&gt; <strong>Credential Type</strong>: Red Hat Satellite 6</li><li><strong>Inventories</strong> -&gt; <strong>Add</strong>, give the inventory a name and <strong>Save</strong></li><li>In the Inventory select the <strong>Sources</strong> -&gt; Add Source -&gt; in the Source box pick Red Hat Satellite 6, and in Credential, pick the set configured in step a above.</li></ol><p>That’s all it takes. Satellite and Ansible Tower now effectively share a brain.</p><p>As the Dynamic Inventory represents all clients in Red Hat Satellite, it <strong>should not</strong> be used directly in Job Templates or Workflows.</p><p>Smart Inventories (filtered versions of the main Dynamic Inventory) should be used by Job and Workflow Templates.</p><p>Filtering used to define Smart Inventories can be based on</p><p>- a parameter set during the provisioning process (and subsequently removed), to perform provisioning steps only on that one system at that part of its provisioning flow.</p><p>- a different parameter can be set once the client is operationalised, to include it in periodic configuration management runs on groups of systems.</p><p>This is done using the following as the Smart Host Filter definition in the Smart Inventory:</p><p><em>variables.icontains:&lt;name of Foreman parameter&gt;</em></p><p>Note that what satisfies this criteria is whether the specified parameter <strong>is defined</strong> in Foreman, rather than its value. Using this method, parameters can be used as flags, determining whether hosts are included in the Smart Inventory that a job runs on, or not.</p><p>The importance of filtering hosts from a Dynamic Inventory rather from just listing them in a regular inventory (or deriving them from other sources, like a Dynamic Inventory from the virtualisation layer) is the key that enables this architecture. When the host is thus pulled from the Satellite 6 Dynamic Inventory, its record brings in</p><ol><li>All the host’s Satellite settings (e.g. Content View, Life Cycle Environment, etc.).</li><li>All the host’s Foreman parameters, as set in Satellite, using either simple (global, per-Organisation/Location/Domain, Hostgroup tree or host “inherit-or-override” logic) or smart (user-specified fact-based logic) parameter values.</li></ol><p>An small sample of useful fields is shown in the example host entry below. <br>Note in particular the foreman_params section at the end, that combines both individual (foreman) parameters applicable to the host and those it inherited from its Hostgroup:</p><pre>foreman:<br>...<br>  content_facet_attributes:<br>    content_view_id: 2<br>    content_view_name: cv-openshift<br>    content_source_id: 1<br>    uuid: e0135ede-59c1-4fd8-a9e3-4957ef339c4d<br>...<br>    lifecycle_environment_name: lce-openshift<br>...<br>    content_source:<br>      url: &#39;<a href="https://infra-satellite.vebate.it:9090&#39;">https://infra-satellite.example.com:9090&#39;</a><br>      id: 1<br>      name: infra-satellite.example.com<br>    upgradable_package_count: 8<br>    lifecycle_environment:<br>      id: 2<br>      name: lce-openshift<br>    id: 1<br>    errata_counts:<br>      bugfix: 0<br>      security: 2<br>      enhancement: 6<br>      total: 8<br>  ip: 192.168.1.159<br>...<br>foreman_facts:<br>...<br>  macaddress: &#39;00:1a:4a:16:01:00&#39;<br>...<br>foreman_params:<br>   ntp_server: 192.168.1.151</pre><p>A playbook can thus refer to an Ansible variable (such as “{{ foreman_params.ntp_server }}” in the example above) sourced from Foreman’s parameters.</p><p>Satellite/Foreman’s individualised information about each system propagates to Ansible playbooks, which Ansible Tower runs. For example, when a single scheduled job will run an SOE playbook on all the hosts in a large fleet, using this method, the playbook on each client will act according to the individual Parameters (Foreman’s term) / Variables (Ansible’s term) for that host.</p><p>A preferred practice may be to avoid modifying third party Ansible roles, but rather to map their internal variables onto Foreman Parameters using group_vars/all file in the Ansible role, eg:</p><pre>ntp_server: “{{ foreman_params.ntp_server }}”</pre><h3>3. Set Up a SOE Job Template and Playbook</h3><p>Ansible Tower offers a powerful capability to separately manage</p><ol><li>What systems a Job Template and its assigned Playbook apply to (defined in the Job Template), and</li><li>What the Job Template does (set in the ansible code making up the Playbook and its Roles, and synchronised from git).</li></ol><p>For example, a playbook can apply an SOE. Its application may need to be performed</p><ol><li>On one specific host as the last part of its provisioning process.</li><li>On all the systems in the fleet every hour.</li></ol><p>While we can set this up as two separate Ansible Tower Projects (hopefully synchronising to the same git repository), a cleaner method is to perform both (related) tasks from the same project, whereby the same playbook will be called by two Job Templates — a scheduled one acting hourly on all operationalised systems in the fleet, and one that is invoked from the provisioning workflow, set to act on just one host.</p><h3>4. Set Up Fancy Provisioning</h3><p>A typical Satellite and Ansible joint provisioning process can, for example, take the form of Ansible Tower Workflow Template(s) with the following shape:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/787/0*DC6S96pXxssCLcD3" /></figure><p>An end-to-end example of the playbooks making up the above workflow, written to accompany this writeup and thoroughly documented inline, can be found <a href="https://github.com/mikishapiro/example-provisioning">here</a>.</p><p>First off, the user will be presented a Survey, asking the relevant information needed to build the client.</p><p>Setting up the survey is dirt simple. Setting it up looks like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/943/0*PSNBf2cftKLXQ7RZ" /></figure><p>And results in this any time the workflow is run:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/879/0*lXYLkqU7RE6rfUfZ" /></figure><p>The playbooks will use the contents of var_hostname and var_hostgroup from the survey to figure out what to do:</p><p>The <strong>first</strong> step will be a git project synchronization step, to ensure Ansible uses the latest playbook code.</p><p>The <strong>second</strong> step will run a Job Template implementing Stage 1 of provisioning — essentially the tasks that happen <em>outside the client</em>.</p><ul><li>The Job Template will run on an <em>empty</em> inventory. An empty inventory has one implied host — localhost (which is the Ansible Tower node), and that is where we want all API calls to originate from.</li><li>This stage will inevitably be specific to Vmware. A provisioning process for another platform — perhaps AWS or bare metal — will need a modified Stage 1.</li></ul><p>This example provisioning process will employ a Vmware template that has a blank disk (e.g. no “Gold Image”), and a suitable Foreman Discovery Image ISO (from the <strong>foreman-discovery-image</strong> rpm available on Satellite, which also provides the foreman-remaster tool to bake in answers for unattended deployments) mounted as a virtual CD.</p><p><strong>Option 1:</strong></p><p>DHCP is available. The FDI image used is a remastered ISO using DHCP for IP settings, also containing the Satellite/Capsule URL.</p><p>The playbook’s first steps will be:</p><ol><li>Use the vmware_guest module to create the VM in “poweredon” mode, and register its json output so the mac address field can be used.</li><li>Using the mac address from the previous step, use the Satellite API to create a discovery rule in Satellite that will provision the client using the user-nominated Hostgroup.</li><li>The second last step will wait for the system to be built and become accessible. The Ansible wait_for_connection module can be used, in combination with a delegate_to: &lt;intended-client-hostname-from-user-survey&gt;.</li></ol><p><strong>Option 2:</strong></p><p>DHCP is <strong>not</strong> available. The default FDI image (requiring the user to log on to the console of the provisioned system and specify the settings required to reach Satellite) will get used.</p><p>The playbook’s first steps will be the same as Option 1 above.<br>Once the VM starts, the user needs to perform a manual step on the VM console.</p><p><strong>Option 3:</strong></p><p>DHCP is <strong>not</strong> available. The playbook can implement the following first steps:</p><ol><li>Create the Host on Satellite using the API.</li><li>Use the Satellite API to download the (very small) host-specific image that has the client’s IP settings baked in.</li><li>Use the Vmware module or API to upload this image to a Vmware datastore.</li><li>Using the Vmware module, create the new VM from a template, in stopped mode.</li><li>Use Vmware module or API, mount the host-specific CD image to the template.</li><li>Using the Vmware module, start the VM and boot from the CD.</li></ol><p>Finally, the last step in Stage 1 will use the Red Hat Satellite API to set a “stage_1_complete_flag” host parameter which marks the system as having completed Stage 1 and ready for Stage 2.</p><p>The <strong>third</strong> workflow stage will re-synchronise the Dynamic Inventory. Though the new client (and any parameters associated with it) is now defined in Satellite, the main Dynamic Inventory in Ansible Tower will not have its record until after this sync step.</p><p>The <strong>fourth</strong> stage in the Workflow will run a provisioning Stage 2 Job Template. This step can be shared by this workflow and others for other platforms (e.g. AWS, bare metal, etc). It will:</p><ul><li>Use a Stage 2 Smart Inventory cut from the now synchronised main Dynamic Inventory, which includes solely the system(s) with the “stage_1_complete_flag” parameter set in the last step of Stage 1.</li><li>The first task in the playbook will remove this parameter, so the system isn’t picked up in subsequent runs, and replace it with a “stage_2_in_progress_flag” parameter.</li><li>The second task in the playbook will either use include (to include playbooks) or include_role (to instantiate roles) to execute its SOE and related payload.</li><li>The last task in the playbook will replace the “stage_2_in_progress_flag” parameter with a “operational_system_under_ansible_control” parameter, which will ensure the system is included in Smart Inventories of subsequent configuration management jobs.</li></ul><p>The workflow is now complete, and the system is operational.</p><h3>In Summary</h3><p>In this architecture, we have demonstrated how to leverage the strengths of both Red Hat Satellite and Ansible Tower, in a solution that delivers best of breed patching, provisioning, configuration management, node classification and workflow automation capabilities, and adheres to the healthy operational practices set out by our starting requirements.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=53c2cd1626f6" width="1" height="1" alt=""><hr><p><a href="https://medium.com/accountable-design/red-hat-satellite-and-ansible-tower-53c2cd1626f6">Red Hat Satellite and Ansible Tower</a> was originally published in <a href="https://medium.com/accountable-design">Accountable Design</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Write Requirements to Defend Decisions]]></title>
            <link>https://medium.com/accountable-design/how-to-write-requirements-to-defend-decisions-379609a992ee?source=rss-c6c055d9e96d------2</link>
            <guid isPermaLink="false">https://medium.com/p/379609a992ee</guid>
            <category><![CDATA[accountable-design]]></category>
            <category><![CDATA[accountable]]></category>
            <category><![CDATA[design]]></category>
            <dc:creator><![CDATA[Miki Shapiro]]></dc:creator>
            <pubDate>Fri, 09 Mar 2018 21:31:21 GMT</pubDate>
            <atom:updated>2020-06-05T05:16:46.100Z</atom:updated>
            <content:encoded><![CDATA[<p>In this piece I will discuss requirements — what they are, why they matter, and how they empower a designer.</p><h3>What Are Requirements?</h3><p>The dictionary doesn’t do a particularly good job of helping me out on this one, so I’m going to offer my own definition.</p><p><strong>Requirements</strong> (in the <em>plural</em>), as used in design, <strong>are</strong> <strong>a written artefact used to capture overarching conditional agreement</strong>, specifically to deem a subsequent proposal as fit for a purpose.</p><p><strong>Requirements</strong> (still in the <em>plural</em>) <strong>are a single list</strong> of conditions or criteria, whereby <strong>if the list in its entirety is satisfied, the customer is bound to agree to the proposal’s fitness</strong>.</p><p>To the language buffs, that’s a pretty uncommon way to use a plural to mean a singular of a related but different thing.</p><p>A requirement (in the <em>singular</em>) is an item on the above list.</p><h3>What Makes A Good Method</h3><p>There are three core distinctions (and several optional ones, like ownership and functional/non-functional which can be added where they add value) which are very good at usefully categorising requirements for designers:</p><ul><li>Hard requirements making a system “<strong>fit for purpose</strong>”, versus soft “<strong>best effort</strong>” requirements that a design is expected to try and meet, but, if not met, do not disqualify the design’s fitness.</li><li><strong>Negotiable</strong> requirements which the customer has the power to retract or amend, versus <strong>non-negotiable</strong> requirements which the customer does not have the authority to withdraw. These are often legal compliance requirements, or internal policy compliance requirements within large organisations.</li><li>Requirements discussed and <strong>agreed to be met</strong>, versus requirements discussed and agreed to <strong>intentionally not be tackled</strong>.</li></ul><h3>MoSCoW</h3><p>An excellent system that can be made to easily communicate all three distinctions above is the <a href="https://en.wikipedia.org/wiki/MoSCoW_method">MoSCoW Method</a> (<strong>M</strong>USTs, <strong>S</strong>HOULDs, <strong>C</strong>OULDs and <strong>W</strong>ON’Ts), or <a href="https://www.ietf.org/rfc/rfc2119.txt">RFC 2119</a> which shares similar wording, albeit with some subtle differences in meaning.</p><p>The way I use them in the real world, to hit all three distinctions above, is:</p><p><strong>MUST</strong> is a fit-for-purpose requirement forced on the customer by an external body, and over which the customer has no control or authority.</p><p><strong>SHOULD</strong> is a fit-for-purpose requirement over which the customer has authority, and should it be required to alter or retract such a requirement, it is within the customer’s authority to modify it.</p><p><strong>COULD</strong> is a best effort requirement — one which a design will make every attempt to meet. Where it differs from SHOULDs and MUSTs is that a design is still fit if a COULD cannot be met. <em>An </em><strong><em>accountable design</em></strong><em> will </em><strong><em>justify</em></strong><em> a decision to not meet a COULD, and </em><strong><em>cite</em></strong><em> competing requirements to justify it.</em></p><p><strong>A COULD carries a secondary design role as a tie-breaker</strong>. When two equally viable options are available, yet one needs to be chosen, a choice is forced, and the rationale, however minor, captured in the form of a COULD requirement, and subsequently cited to justify the design choice.</p><p><strong>WON’T</strong> means deliberate omission. “<strong>We’ve thought about it and decided not to</strong>” (as opposed to “Whoops, we totally forgot”). An often overlooked yet crucially important form of requirement, a WON’T allows the customer to set hard boundaries on scope on one hand (preventing, for example, a paid designer from spending time on something regarded as unnecessary), yet in doing so, the responsibility for the omission be correctly attributed to customer and not fall on the designer, should the omission have consequences.</p><h3>Make It Easy For People To Use Your Good Method</h3><p>A good method that crystallises the distinctions isn’t enough. There are several more things you need do to make it as easy as possible to consume — for the designer, for decision maker and operational customer stakeholders and also for the technology vendor writing a lookup table.</p><p>Your requirements also need to be:</p><p><strong>Itemised</strong>. Rows in a table. When requirements are presented as a single 1,500 word (or 15,000 word) blob of text, at the beginning of some document, it is hard to pinpoint which of those words are the reason for a decision, or, more critically, which of our decisions are impacted if we iterate with modified or retracted requirements. You have to re-evaluate all your decisions for impact, and end up in documentation hell. And avoiding baking in solution mode in such a text is virtually impossible.</p><p>Furthermore, after agreement was reached on such a text as a whole, nobody will ever want to change a single letter in it, for fear of invalidating the agreement reached, and having to go through the whole painful process again.</p><p>A second reason to break them down into an itemised list is keeping your audience(s) understanding and owning the bits they understand and can talk to, without contaminating their requirements with other items they cannot speak to.</p><p>The senior decision maker will have put the 3 business requirements at the top of the list. That’s why you are there. The operational team will have put another 20 technical requirements underneath. Fuse them together, and everyone glazes over. Keep them separate, and all the heads in the room are nodding, even with a mixed audience — the senior decision maker is happy because their 3 business requirements are at the top of the list and they can map them straight to the targets motivating them. The ops team are happy because the subsequent items demonstrate understanding of the constraints of the real world, and they are empowered to steer and shape those requirements. It is the same for stakeholders who force compliance requirements, operational handover groups with their requirements and so forth.</p><p><strong>Citable</strong>. Enumerate them, and keep the enumeration consistent. Requirements are power to wield the customer’s authority to justify decisions. It’s <strong>a designer’s version of “privilege escalation”.</strong> But it only works if you can get agreement on a customer’s words in writing, put an altar of authority underneath the words — the requirements to the design are such an altar — and cite them accurately and prudently when justifying decisions.</p><p>One (or more) requirement needs to be cited for every decision we write down. Whatever the format, an accountable design will capture, even at a high level, what options were considered, and cite one or more requirements as the reason for the choice.</p><p><em>If multiple viable options exist, and the requirements do not tip the scales on either… </em><strong><em>you forgot to list a requirement.</em></strong><em> Pick the right answer. Say why. That reason is the missing requirement (perhaps a COULD). Add it and cite it.</em></p><p><strong>Changeable</strong>. Design is<strong> </strong>an<strong> iterative process of discovery, understanding, creativity and compromise</strong>. The more complex, the more likely that the design group will feature people who are experts in specialised subject matter fields, or in the nuanced world of one stakeholder or another, and whose needs and inputs will need to sometimes be weighed against one another.</p><p>Final, well articulated requirements that hit the mark from their first draft have been presented to a design team, on day one, on a silver platter, precisely nowhere, ever. It doesn’t happen when you design a CI/CD pipeline, and it doesn’t happen when you design an airliner either.</p><p>Rather, final, well articulated requirements to what will become the solution are forged. Some are born in unworkable states, and get iteratively refined through parallel problem and solution conversations until they become workable.</p><h3>Talking It Through With Stakeholders</h3><p>First, <strong>this should not be assumed knowledge</strong>. Nobody is born knowing this. I find it extremely helpful to spend 5 minutes agreeing with each group I set out to do design work with, that we will write down their needs in the form of requirements, not bullet points or notes or freeform text. There is often a sharp intake of breath when people think you’re about to ask they present a thoroughly documented set of requirements, which they prepared ahead of time. There’s an audible sigh of relief when you say it’s ok, part of the design work is to help them put it together, write it down for them during the design period, and ultimately ask them for signoff.</p><p>Second, I review with them quickly the four MoSCoW types, what each means, and how we will use them. This gives them an opportunity to customise the method to capture additional distinctions they find important, or to comply with existing practices within their organisation.</p><p>Last, a few days later, once most of the requirements have been understood and written down, a quick diagram or whiteboard photo and a high level solution approach agreed upon, I circulate an email to the customer with all the requirements in a table, and ask them eyeball, correct if I got anything wrong, and when all is good, respond with a simple written “Yes”.</p><p>It is my charge to make sure every decision in the design has at least one requirement I can cite, to put spine behind the choice made. It is never my whim or preference.</p><p>Even citing “leading practice” should in my view not be… design leading practice, pun half-intended. If something is leading practice, it probably has firm reason behind it. Why it is so and when it applies should be understood. Following leading practice is a <strong>decision</strong> you may wish to document, in the design decisions, not in the requirements. The reason to follow the practice is the requirement.</p><p>When all decisions made can be traced back to customer needs, the design is and will remain, fully accountable.</p><p>In the next piece I will discuss bringing accountable design into the world of AGILE, where documentation is secondary priority to a working product. I will discuss <strong>lean</strong> design documents — what content classes can be found in designs, which have more value, and which have less.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=379609a992ee" width="1" height="1" alt=""><hr><p><a href="https://medium.com/accountable-design/how-to-write-requirements-to-defend-decisions-379609a992ee">How to Write Requirements to Defend Decisions</a> was originally published in <a href="https://medium.com/accountable-design">Accountable Design</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Simplifying Design with Lookup Tables]]></title>
            <link>https://medium.com/accountable-design/simplifying-design-with-lookup-tables-3915bcc84cc4?source=rss-c6c055d9e96d------2</link>
            <guid isPermaLink="false">https://medium.com/p/3915bcc84cc4</guid>
            <category><![CDATA[design]]></category>
            <dc:creator><![CDATA[Miki Shapiro]]></dc:creator>
            <pubDate>Fri, 09 Mar 2018 21:30:12 GMT</pubDate>
            <atom:updated>2018-03-09T21:30:12.892Z</atom:updated>
            <content:encoded><![CDATA[<h3>Lookup Tables</h3><p>In this second article in a series discussing the fundamentals of Accountable Design, I will provide an overview of using <strong>lookup tables</strong>. <br>An excellent way to communicate, justify and document leading practice, particularly when such practice is only advised under specific conditions, is through the use of a lookup table.</p><p>Why is this good?</p><ul><li>To<strong> efficiently context and recycle knowledge</strong> from tested solutions and past experiences.</li><li>To <strong>help human designers avoid overlooking a better option</strong> to any one problem.</li><li><strong>Requirements are sometimes missed</strong> by both designers and customers.</li><li>To <strong>raise the accountability </strong>bar on the designs we write.</li></ul><p>A lookup table is a design aid not specific to the solution, and should cover the primary areas of a technology set where a designer needs to make decisions. It should name the conditions, and recommend courses of action.</p><p>For example, the following table describes 4 requirements a customer may have regarding the solution’s high availability:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/661/1*jiulcQ9n6aNwLvmmmHgAOw.png" /></figure><p>Obviously, not all requirements will be relevant to each customer. Some are more extreme versions of others. Yet others may be mutually exclusive.</p><p>I’d like to come back to the “Why is this good?” I stated earlier.</p><ul><li><em>To efficiently context and recycle knowledge from tested solutions and of past experiences.</em></li></ul><p><strong>A lookup table is a field index</strong> to knowledge about a technology, possibly spread across multiple information management systems (manuals, wikis, knowledge base articles, bugs, tickets, web articles, etc.)</p><ul><li><em>To help human designers avoid overlooking better suited options to any one problem.</em></li></ul><p><strong>A centrally managed lookup table ensures designers are aware of all options </strong>available.</p><ul><li><em>Requirements are sometimes missed by both designers and customers.</em></li></ul><p><strong>A lookup table, if traversed methodically, also functions as a checklist</strong>, performing the same role as a pre-flight check-list, reducing the likelihood of oversight.</p><ul><li><em>To raise the accountability bar of the designs we write.</em></li></ul><p>Rooting design decisions in best-of-breed, situation-dependent, leading practice advertised by a trusted source makes for accountable design.</p><p>In summary, lookup tables are a powerful tool to manage both knowledge and design quality.</p><h3>Start Anywhere</h3><p>Perhaps one of the nicest things about lookup tables is that they can originate and be useful at any level, and can be driven both top-down or bottom-up in an organisation.</p><p><strong>An</strong> <strong>individual designer</strong> can keep track of their personal accrued experience using itemised requirement/solution lookup tables.</p><p><strong>A team of professionals</strong> can likewise track, share and correctly apply their situationally-specific accrued experiences needing little from their employer, using a lookup table on a central location such as an internal wiki.</p><p><strong>A business unit specialising in a specific technology</strong>, whose product requires some level of local solution design work and/or integration into different types of heterogenous environments, can share a top-down lookup table to not only publish what it sees as leading practice, but to help its audience apply the correct practices in the correct situations. It can sometimes make sense to have have two versions, a public version with customer-facing messaging, and a private one with extra options worthy of internal circulation.</p><p><strong>A technology community</strong>, such as an open source project, can likewise advertise to its audience how and in what circumstances its products can be used.</p><p>In the next article I will take a closer look at <strong>requirements</strong>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3915bcc84cc4" width="1" height="1" alt=""><hr><p><a href="https://medium.com/accountable-design/simplifying-design-with-lookup-tables-3915bcc84cc4">Simplifying Design with Lookup Tables</a> was originally published in <a href="https://medium.com/accountable-design">Accountable Design</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>