<?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 Enrique Somoza, D.Sc. on Medium]]></title>
        <description><![CDATA[Stories by Enrique Somoza, D.Sc. on Medium]]></description>
        <link>https://medium.com/@enriquesomoza?source=rss-75f0c3cd9f0b------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*ZR03Y4OW_nbpXF_yfQzumg.jpeg</url>
            <title>Stories by Enrique Somoza, D.Sc. on Medium</title>
            <link>https://medium.com/@enriquesomoza?source=rss-75f0c3cd9f0b------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Mon, 16 Nov 2026 20:42:57 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@enriquesomoza/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[Five Books Product Managers Should Be Reading in 2027]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://enriquesomoza.medium.com/five-books-product-managers-should-be-reading-in-2027-24caf269cf83?source=rss-75f0c3cd9f0b------2"><img src="https://cdn-images-1.medium.com/max/2600/1*2gnWNzioDngtJOJvMZGhkQ.jpeg" width="5555"></a></p><p class="medium-feed-snippet">Why judgment, trust, and accountability define the reading list this year</p><p class="medium-feed-link"><a href="https://enriquesomoza.medium.com/five-books-product-managers-should-be-reading-in-2027-24caf269cf83?source=rss-75f0c3cd9f0b------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://enriquesomoza.medium.com/five-books-product-managers-should-be-reading-in-2027-24caf269cf83?source=rss-75f0c3cd9f0b------2</link>
            <guid isPermaLink="false">https://medium.com/p/24caf269cf83</guid>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[product-design]]></category>
            <category><![CDATA[product-development]]></category>
            <dc:creator><![CDATA[Enrique Somoza, D.Sc.]]></dc:creator>
            <pubDate>Thu, 05 Nov 2026 16:36:05 GMT</pubDate>
            <atom:updated>2026-11-05T16:36:05.673Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Key Tag 38696 Takes Over the Root This Weekend]]></title>
            <link>https://meetcyber.net/key-tag-38696-takes-over-the-root-this-weekend-2359f48b7e0a?source=rss-75f0c3cd9f0b------2</link>
            <guid isPermaLink="false">https://medium.com/p/2359f48b7e0a</guid>
            <category><![CDATA[dns]]></category>
            <category><![CDATA[security]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[Enrique Somoza, D.Sc.]]></dc:creator>
            <pubDate>Wed, 07 Oct 2026 21:45:34 GMT</pubDate>
            <atom:updated>2026-10-07T21:45:34.935Z</atom:updated>
            <content:encoded><![CDATA[<h4>What actually happens during the 2026 DNSSEC root KSK rollover, and what breaks when a resolver misses it</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*n1PZ-S5jv_PlFsJg1IFEVA.png" /></figure><p>On October 11, something very small will change at the top of the DNS.</p><p>One key will start signing.</p><p>Another will stop.</p><p>For most of the Internet, absolutely nothing should happen.</p><p>That is the point.</p><p>But underneath that uneventful change is one of the more interesting distributed security operations we perform on the Internet.</p><p>The DNS root Key Signing Key is <a href="https://www.icann.org/resources/pages/ksk-rollover-en">rolling over</a>.</p><p><strong>The incoming key is KSK-2024.</strong></p><p><strong>Its key tag is 38696.</strong></p><p>The outgoing key, KSK-2017, has key tag 20326 and has been signing the root DNSKEY RRset since October 11, 2018. On October 11, 2026, KSK-2024 becomes the active signing key.</p><p>If you operate a DNSSEC-validating recursive resolver, this is worth understanding at a level deeper than: “Make sure 38696 is in your trust-anchor file.”</p><p>Because the interesting part of a KSK rollover isn’t really changing the key.</p><p><strong>It’s changing trust.</strong></p><h3>Start at the top of the chain</h3><p>DNSSEC validation ultimately needs a starting point. For the global DNS, that starting point is the root trust anchor.</p><pre>Root KSK<br>↓ signs<br>. DNSKEY RRset<br>↓ authenticates root ZSK<br>Root zone records<br>↓ DS for .com<br>.com DNSKEY<br>↓ DS for example.com<br>example.com DNSKEY<br>↓<br>www.example.com</pre><p>There is an important distinction here. People often describe the KSK as “the key that signs the ZSK.” That’s a useful simplification, but it isn’t literally how DNSSEC works.</p><p>The KSK signs the DNSKEY RRset. The ZSK is a member of that RRset. So validating the KSK’s RRSIG over the DNSKEY RRset authenticates the ZSK, which can then authenticate the other signed RRsets in the root zone.</p><h3>Meet 38696</h3><p>KSK-2024 is an RSA/SHA-256 DNSSEC key using algorithm 8.</p><pre>. 172800 IN DNSKEY 257 3 8 …</pre><p>The 257 flags value is ZONE (256) + SEP (1). The protocol value is 3. Algorithm 8 is RSASHA256.</p><p>When the DNSKEY key-tag algorithm is applied to KSK-2024, the result is 38696 (0x9728). IANA identifies the key as KSK-2024, label Kmyv6jo.</p><p>The predecessor is KSK-2017, key tag 20326, also algorithm 8.</p><h3>October 11 isn’t when 38696 appears</h3><p>38696 is not suddenly appearing this weekend. The successor was generated in April 2024, published well ahead of operational use, and introduced into the root DNSKEY set in 2025.</p><p>That long overlap exists because validating resolvers need time to learn: “I trust this key too.” That’s where <a href="https://www.rfc-editor.org/rfc/rfc5011.html">RFC 5011</a> enters the story.</p><h3>RFC 5011 is doing the real work</h3><p>RFC 5011 defines automated DNSSEC trust-anchor maintenance. A resolver that already trusts KSK-2017 can observe KSK-2024 inside the securely validated root DNSKEY RRset, but it does not immediately trust the new key.</p><pre> KSK-2017: TRUSTED<br> ↓ validates root DNSKEY RRset<br> KSK-2024 appears<br> ↓<br> ADD-PENDING<br> ↓ continuously observed for hold-down period<br> VALID / TRUSTED</pre><p>The add hold-down time is nominally 30 days. <a href="https://www.iana.org/dnssec/files">IANA’s operational schedule</a> gives January 11, 2025 as publication and February 10, 2025 as the point at which RFC 5011-compliant resolvers should begin trusting the successor.</p><p>Then IANA waits. That long pre-publication period creates an enormous operational safety margin.</p><h3>What actually changes October 11</h3><p>Before the rollover, both generations of KSK material are present during the transition. A properly updated validating resolver should already trust both 20326 and 38696.</p><p>On October 11, signing responsibility changes: KSK-2024 becomes the key used to sign the root DNSKEY RRset and KSK-2017 stops performing that active signing role.</p><p>The important change is not that 38696 suddenly exists. It is that the validation path now depends on 38696.</p><h3>What happens if your resolver doesn’t trust 38696?</h3><p>If a validating resolver still trusts only 20326 after the signing transition, it cannot establish the expected trust path through the new active root KSK.</p><p>For users, the symptom is likely to be SERVFAIL. Network connectivity can remain perfectly healthy while DNSSEC-validating resolution fails. It may not appear at the moment of the change, either. The root DNSKEY RRset has a 48-hour TTL, so a resolver that validated and cached it just before the rollover can keep answering for up to two days before validation starts to fail.</p><p><strong>Your network isn’t down.</strong> Your resolver no longer has the trust state it needs to validate the DNS.</p><h3>Seeing 38696 isn’t the same as trusting 38696</h3><p>dig . DNSKEY +dnssec</p><p>Seeing KSK-2024 in the root DNSKEY RRset proves that the root publishes it. It does not prove that your validating resolver has accepted it as a trust anchor.</p><p>Operators should inspect the resolver’s actual trust-anchor state. In BIND, bind.keys (or the keys compiled into named) only seeds the initial trust anchors. The learned RFC 5011 state lives in the managed-keys database, and rndc managed-keys status or rndc secroots shows what named trusts right now. Other resolver implementations keep it in root.key or root.keys-style trust-anchor files, depending on packaging and configuration.</p><p>The state that matters is not “38696 is visible.” It is “38696 is trusted.”</p><h3>Why automatic rollover can still fail</h3><p>RFC 5011 automates the process, but automation can fail operationally: trust-anchor maintenance may be disabled, state files may not be writable, container images may repeatedly restore stale state, configuration management may overwrite learned state, or a resolver may have been offline during critical observation periods.</p><p>Running a modern resolver version is not equivalent to proving that 38696 is trusted.</p><p><a href="https://www.icann.org/en/blogs/details/preparing-for-the-root-zone-ksk-rollover-what-you-need-to-know-27-07-2026-en">ICANN has reported</a> that more than 95% of reporting resolvers have recognized and adopted KSK-2024. That is reassuring ecosystem telemetry, but it is not evidence about any particular resolver.</p><h3>The old key doesn’t disappear Sunday</h3><p>· <strong>April 2024:</strong> KSK-2024 generated</p><p>· <strong>2025:</strong> KSK-2024 pre-published</p><p>· <strong>October 11, 2026:</strong> KSK-2024 takes over signing</p><p>· <strong>January 2027:</strong> KSK-2017 revocation phase</p><p>· <strong>Later 2027:</strong> old key material retired</p><p>A trust-anchor rollover is not “old key off, new key on.” It is a staged distributed-state transition.</p><h3>And that’s the part I find fascinating</h3><p>We tend to describe DNSSEC as cryptography: signatures, keys, DS records, algorithms and hashes.</p><p>But the root KSK rollover is not primarily difficult because RSA/SHA-256 is difficult. The hard part is coordinating trust state across independently operated systems all over the Internet.</p><p>One organization publishes a new key. Millions of independently operated recursive resolvers have to observe it. Each resolver maintains local state. That state has to survive restarts, containers, package upgrades, configuration management, network outages, software differences and operational mistakes.</p><p>Then the global system changes which key it depends on. There is no central database that can say every resolver is ready. DNS does not work that way.</p><p><strong>That’s distributed systems engineering disguised as key management.</strong></p><h3>What I’d watch this weekend</h3><p>· SERVFAIL rate</p><p>· DNSSEC BOGUS results</p><p>· Root DNSKEY refreshes</p><p>· Validation failures</p><p>· Trust-anchor state</p><p>· Extended DNS Errors</p><p>· TCP fallback</p><p>· Differences between validating and non-validating resolution</p><p>I would keep watching through Tuesday, October 13, because cached root keys can hide a problem for up to 48 hours.</p><p>And I would verify one thing before October 11: 38696 = TRUSTED. Not merely visible. Not probably installed. Trusted.</p><h3>The root of trust is still operational infrastructure</h3><p>Post-quantum DNSSEC is forcing us to rethink the future of the trust chain. The KSK-2024 rollover is a reminder that we still have to operate today’s trust system correctly.</p><p>KSK-2024 uses the same RSA/SHA-256 algorithm as its predecessor. There is no dramatic new cryptography, no new DNS protocol, and no new validation model. Yet changing one key requires years of preparation.</p><p>Because DNSSEC’s real challenge has never just been generating signatures. It’s maintaining a globally consistent chain of trust across infrastructure nobody centrally controls.</p><p>On October 11, 20326 stops being the active signing KSK. 38696 takes over.</p><p>If everything works, most people won’t notice.</p><p>That’s not because nothing happened. It’s because the preparation worked.</p><p>And for something sitting at the root of DNSSEC’s global chain of trust, boring is exactly what we want.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2359f48b7e0a" width="1" height="1" alt=""><hr><p><a href="https://meetcyber.net/key-tag-38696-takes-over-the-root-this-weekend-2359f48b7e0a">Key Tag 38696 Takes Over the Root This Weekend</a> was originally published in <a href="https://meetcyber.net/">MeetCyber</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[We’re Measuring DNS Attacks From the Wrong Side]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://meetcyber.net/were-measuring-dns-attacks-from-the-wrong-side-a1e862a475b2?source=rss-75f0c3cd9f0b------2"><img src="https://cdn-images-1.medium.com/max/1251/1*4Htn6hcJLIgRjQ64taSELw.png" width="1251"></a></p><p class="medium-feed-snippet">Query volume tells us what entered a resolver. It doesn&#x2019;t tell us what the resolver was forced to do.</p><p class="medium-feed-link"><a href="https://meetcyber.net/were-measuring-dns-attacks-from-the-wrong-side-a1e862a475b2?source=rss-75f0c3cd9f0b------2">Continue reading on MeetCyber »</a></p></div>]]></description>
            <link>https://meetcyber.net/were-measuring-dns-attacks-from-the-wrong-side-a1e862a475b2?source=rss-75f0c3cd9f0b------2</link>
            <guid isPermaLink="false">https://medium.com/p/a1e862a475b2</guid>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[cloud-security]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[data-security]]></category>
            <dc:creator><![CDATA[Enrique Somoza, D.Sc.]]></dc:creator>
            <pubDate>Tue, 06 Oct 2026 14:28:04 GMT</pubDate>
            <atom:updated>2026-10-06T14:28:04.064Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[1,600 Applications Later: What Will Decide Which New TLDs Last]]></title>
            <link>https://enriquesomoza.medium.com/1-600-applications-later-what-will-decide-which-new-tlds-last-7d0310839b39?source=rss-75f0c3cd9f0b------2</link>
            <guid isPermaLink="false">https://medium.com/p/7d0310839b39</guid>
            <category><![CDATA[domain-names]]></category>
            <category><![CDATA[dns]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[internet-governance]]></category>
            <category><![CDATA[icann]]></category>
            <dc:creator><![CDATA[Enrique Somoza, D.Sc.]]></dc:creator>
            <pubDate>Thu, 10 Sep 2026 12:44:10 GMT</pubDate>
            <atom:updated>2026-09-10T12:44:10.368Z</atom:updated>
            <content:encoded><![CDATA[<h4>ICANN closed the 2026 application window on August 12 with more than 1,600 primary applications. That is fewer than the 1,930 submitted in 2012, but the number is not the most interesting part of this round.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*F87SvZtc4UqVsCEAJtzSDw.png" /></figure><p>The process has changed. Private resolution of contention sets is gone. Replacement strings are new. And every applicant had to identify a registry service provider that ICANN had already evaluated.</p><p>Over the next few months, most of the attention will be on Reveal Day, contention sets and, eventually, auctions. That is understandable. It is also not what will determine whether these new TLDs succeed.</p><p>That happens after delegation.</p><p>It happens when the DNS is under attack. When the registry goes through its first DNSSEC key rollover. When a registrar promotion attracts the wrong kind of registrations. And when the team that fought to launch the string has to decide who will own it for the next ten years.</p><p>I spent six years managing registry products for .co, .us, .biz and .au, followed by another six leading product for an authoritative DNS platform used by TLD operators. Earlier this year, I also worked through an RSP evaluation application. Based on that experience, here is what I think every new registry operator — and every product team building for one — should be thinking about now.</p><h3>A string is not a product</h3><p>The 2012 round taught us this the hard way.</p><p>Many TLDs launched with a plan built almost entirely around the opening sequence: sunrise, landrush, general availability and a registrar push. Once the launch was over, some never found a durable reason to exist. Dozens of brand TLDs were eventually returned to ICANN because no one inside the company truly owned what the TLD was supposed to become.</p><p>A TLD needs a product lifecycle like anything else. It starts with the concept and business case, moves through application, evaluation and launch, and then enters the much longer part of its life: pricing changes, registrar relationships, policy updates, abuse management, technical migrations and, eventually, renewal.</p><p>The registries that performed well after 2012 were prepared for that long middle — not just the launch.</p><p>This also cannot sit entirely with the product team. Pricing affects both revenue and abuse. Eligibility rules touch legal, policy and marketing. Registrar programs require account management, operations and support. If those groups only come together for launch, the TLD will eventually lose direction.</p><h3>The RSP list is a starting point</h3><p>The Registry Service Provider Evaluation Program is one of the strongest changes in the new round.</p><p>In 2012, technical capability was assessed application by application, creating hundreds of evaluations that covered much of the same ground. This time, providers are evaluated once across core registry functions, DNS, DNSSEC and related services, and applicants select from the approved list.</p><p>That is a meaningful improvement. But an evaluated provider is not the same thing as the right provider for your TLD.</p><p>The list tells an applicant that a provider met ICANN’s requirements at the time of evaluation. It does not tell you how the provider will respond during a sustained DDoS attack against your nameservers. It does not tell you how quickly a zone change will publish after a registrar suspends a phishing domain. It does not tell you how the provider intends to manage DNSSEC algorithm transitions over the next decade — or how difficult it will be to leave if the relationship no longer works.</p><p>Having worked through those evaluation questions, I think they are good questions. Registry operators should still ask them again, in their own words, before signing a contract.</p><h3>DNS is the infrastructure everyone takes for granted, until it fails</h3><p>The base registry agreement has long set the expectation at zero minutes of total DNS service downtime. That is appropriate. A TLD sits above every domain registered beneath it. If its authoritative DNS fails, websites, mail servers and APIs across the entire extension can disappear with it.</p><p>TLD nameservers are also attractive targets.</p><p>Random-subdomain attacks, often called DNS water torture, generate large volumes of queries for names that do not exist. Because recursive resolvers cannot rely on a cached answer for each new name, much of that traffic reaches the authoritative servers. I have spent a significant part of my career working on how to detect these attacks early. The economics remain stubbornly simple: they are cheap to launch and expensive to absorb.</p><p>A smaller TLD is not protected simply because it has fewer domains. If anything, an operator with limited capacity or a single anycast provider may have less room for error than a large, established registry.</p><p>DNS capacity, attack resilience and provider diversity therefore belong in the business case. They are not technical details to work out after delegation.</p><h3>DNSSEC failures arrive all at once</h3><p>Every new gTLD must be signed. Signing the zone on day one is the easy part.</p><p>The real risk comes later: a signature expires without anyone noticing, a key rollover fails, or the DS record in the root no longer matches the key used to sign the zone. At the TLD level, a DNSSEC failure is not a minor incident. Validating resolvers return SERVFAIL for every affected name below it. To users, the entire extension appears to be gone.</p><p>The safeguards are not exotic. Automate key management and rollovers so they do not depend on someone remembering a date. Use modern algorithms that keep responses smaller and signing efficient. Monitor the zone from the outside, from the same perspective as a validating resolver. And begin planning now for future algorithm changes, including the much larger signatures expected in a post-quantum world.</p><p>A registry agreement lasts for years. The algorithm a TLD launches with is unlikely to be the one it uses forever.</p><h3>Abuse belongs on the business dashboard</h3><p>Since April 2024, registries and registrars have had a contractual obligation to act on DNS abuse, including malware, botnets, phishing, pharming and spam when it is used to deliver those threats.</p><p>ICANN reported in June that it had launched nearly 530 investigations under those requirements during the first two years and resolved more than 480. About two-thirds resulted in action by the registry or registrar, contributing to the mitigation of more than 25,000 abusive domains. ICANN is also developing a proactive enforcement framework tied to abuse indicators, meaning the worst-performing operators may draw attention before anyone files a complaint.</p><p>But the contractual risk is only part of the story. Reputation damage can cost a new TLD far more than a compliance notice.</p><p>When abuse becomes concentrated in an extension, security vendors, mail providers and network operators begin treating the entire TLD as suspicious. Legitimate registrants get caught in filters. Registrars become less enthusiastic about promoting it. Growth slows.</p><p>Aggressive first-year discounts are a common source of this problem. Low prices can produce impressive registration numbers, but some of that volume may be exactly the kind an operator does not want. A registration is not valuable if it creates more cost, risk and reputational damage than revenue.</p><p>That is why abuse rates should sit on the same dashboard as registrations, renewals and revenue. They should be measured by registrar, promotion and acquisition channel. Pricing, eligibility rules and registrar onboarding should all be reviewed with that data in the room.</p><h3>What I would put on the roadmap now</h3><p>If I were building the product plan for a TLD coming out of this round — or for a registry platform serving many of them — I would start with six things:</p><ul><li>A post-delegation operating plan with named owners, not just a launch plan.</li><li>An RSP review that goes beyond approval status and examines attack history, zone-publication latency, DNSSEC operations and exit terms.</li><li>DNS capacity and provider diversity included in the financial model from the beginning.</li><li>Automated DNSSEC operations, outside-in monitoring and a written plan for future algorithm changes.</li><li>Abuse metrics segmented by registrar and promotion, reviewed alongside revenue and renewals.</li><li>Registrar training that covers abuse response and operational expectations, not only onboarding and sales.</li></ul><p>The 2026 round will add hundreds of new TLDs to the root. Ten years from now, the ones still growing will not necessarily be the strings that drew the most attention on Reveal Day or commanded the highest auction price.</p><p>They will be the ones whose operators treated them as long-lived infrastructure products from the beginning.</p><p>The string gets you into the root. Operating it well is what makes it last.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7d0310839b39" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[When DNS Caches Become Distributed State]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://meetcyber.net/when-dns-caches-become-distributed-state-38d71799dfad?source=rss-75f0c3cd9f0b------2"><img src="https://cdn-images-1.medium.com/max/1320/1*NCY7B8a-ae3PgPaqzOK6yw@2x.jpeg" width="1320"></a></p><p class="medium-feed-snippet">Sharing resolver caches could make DNS faster and cheaper. </p><p class="medium-feed-link"><a href="https://meetcyber.net/when-dns-caches-become-distributed-state-38d71799dfad?source=rss-75f0c3cd9f0b------2">Continue reading on MeetCyber »</a></p></div>]]></description>
            <link>https://meetcyber.net/when-dns-caches-become-distributed-state-38d71799dfad?source=rss-75f0c3cd9f0b------2</link>
            <guid isPermaLink="false">https://medium.com/p/38d71799dfad</guid>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[networking]]></category>
            <category><![CDATA[dns]]></category>
            <category><![CDATA[productivity]]></category>
            <dc:creator><![CDATA[Enrique Somoza, D.Sc.]]></dc:creator>
            <pubDate>Fri, 28 Aug 2026 18:16:27 GMT</pubDate>
            <atom:updated>2026-08-28T18:16:27.575Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[DNS Attacks Aren’t Just About Query Volume Anymore]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://meetcyber.net/dns-attacks-arent-just-about-query-volume-anymore-c6a1c3c0d532?source=rss-75f0c3cd9f0b------2"><img src="https://cdn-images-1.medium.com/max/2600/1*fij34v6zCGnUxYY6eL6_Gw.png" width="2794"></a></p><p class="medium-feed-snippet">New research on DNS self-amplification makes me think we&#x2019;ve been watching the wrong side of the resolver.</p><p class="medium-feed-link"><a href="https://meetcyber.net/dns-attacks-arent-just-about-query-volume-anymore-c6a1c3c0d532?source=rss-75f0c3cd9f0b------2">Continue reading on MeetCyber »</a></p></div>]]></description>
            <link>https://meetcyber.net/dns-attacks-arent-just-about-query-volume-anymore-c6a1c3c0d532?source=rss-75f0c3cd9f0b------2</link>
            <guid isPermaLink="false">https://medium.com/p/c6a1c3c0d532</guid>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[dns]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <dc:creator><![CDATA[Enrique Somoza, D.Sc.]]></dc:creator>
            <pubDate>Fri, 07 Aug 2026 21:59:00 GMT</pubDate>
            <atom:updated>2026-08-10T07:32:31.394Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[When UX Is Reliability Engineering: What Designing for Infrastructure Taught Me About Risk]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://meetcyber.net/when-ux-is-reliability-engineering-what-designing-for-infrastructure-taught-me-about-risk-af6ff21dbcee?source=rss-75f0c3cd9f0b------2"><img src="https://cdn-images-1.medium.com/max/1767/1*FfyKyST9MwqyEAmKLyrMFQ.png" width="1767"></a></p><p class="medium-feed-snippet">When UX Is Reliability Engineering: What Designing for Infrastructure Taught Me About Risk</p><p class="medium-feed-link"><a href="https://meetcyber.net/when-ux-is-reliability-engineering-what-designing-for-infrastructure-taught-me-about-risk-af6ff21dbcee?source=rss-75f0c3cd9f0b------2">Continue reading on MeetCyber »</a></p></div>]]></description>
            <link>https://meetcyber.net/when-ux-is-reliability-engineering-what-designing-for-infrastructure-taught-me-about-risk-af6ff21dbcee?source=rss-75f0c3cd9f0b------2</link>
            <guid isPermaLink="false">https://medium.com/p/af6ff21dbcee</guid>
            <category><![CDATA[ux]]></category>
            <category><![CDATA[design]]></category>
            <category><![CDATA[product-design]]></category>
            <category><![CDATA[infrastructure]]></category>
            <category><![CDATA[ux-design]]></category>
            <dc:creator><![CDATA[Enrique Somoza, D.Sc.]]></dc:creator>
            <pubDate>Wed, 29 Jul 2026 21:28:39 GMT</pubDate>
            <atom:updated>2026-07-29T21:28:39.411Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[The Best DNS Products Don’t Add Features. They Remove Operational Risk.]]></title>
            <link>https://blog.stackademic.com/the-best-dns-products-dont-add-features-they-remove-operational-risk-9179c9b57fc6?source=rss-75f0c3cd9f0b------2</link>
            <guid isPermaLink="false">https://medium.com/p/9179c9b57fc6</guid>
            <category><![CDATA[dns]]></category>
            <category><![CDATA[risk-management]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[product-management]]></category>
            <dc:creator><![CDATA[Enrique Somoza, D.Sc.]]></dc:creator>
            <pubDate>Wed, 15 Jul 2026 10:36:32 GMT</pubDate>
            <atom:updated>2026-07-15T10:36:32.983Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*U0-Q3hd0Jj91xurJvZY1XQ.png" /></figure><p>If you look at the postmortems of the most damaging DNS outages of the last decade, a pattern emerges that should make every product leader in this space uncomfortable: almost none of them were protocol failures. The protocol did exactly what it was told. The specifications held. What failed was the operational layer around DNS, the humans, the tooling, the change processes, and the assumptions baked into all three.</p><p>That pattern has convinced me of something I now treat as a product thesis: the best DNS products are not the ones with the longest feature lists. They are the ones that systematically remove operational risk. Everything else is noise.</p><h3>DNS outages are operational failures wearing protocol costumes</h3><p>When a major DNS incident hits the news, the headline usually blames the technology. “DNSSEC misconfiguration takes down service.” “Expired signatures break resolution for an entire country-code TLD.” Read the actual postmortems, though, and the story is almost always the same: a change was made, the change was valid syntax, the system accepted it, and the blast radius was not understood until resolvers around the world started returning SERVFAIL.</p><p>DNS is unforgiving in a way that few other infrastructure layers are. There is no canary deployment for a delegation change. There is no instant rollback when negative caching has already propagated a bad answer across millions of resolvers. TTLs mean that your mistake outlives your fix. A misapplied DS record or a botched key rollover does not fail loudly at deploy time. It fails quietly, globally, and on a delay.</p><p>This is why I keep coming back to the same conclusion: the protocol is not the risk surface. The change is. And products that only expose the protocol, without shaping how changes happen, are shipping the risk directly to their customers.</p><h3>From CLI and APIs to guided operational workflows</h3><p>The first generation of DNS tooling was the zone file and a text editor. The second was the API: programmatic, scriptable, and a genuine leap forward for automation. But APIs have a property that product teams do not talk about enough. An API will let you do anything, including the catastrophic thing, with equal enthusiasm and zero context.</p><p>The evolution now underway across the industry is a third stage: guided operational workflows. Not wizards that dumb things down, but interfaces that encode operational knowledge into the change path itself. What does this look like in practice?</p><ul><li>Surfacing the consequences of a change before it is applied, not after resolvers have cached it.</li><li>Grouping related settings together so an operator configuring rate limiting also sees negative caching behavior, because those decisions interact.</li><li>Making previously API-only settings visible, because a setting nobody can see is a setting nobody audits.</li><li>Providing in-context explanations of concepts like TTL and negative caching at the exact moment an operator is making a decision that depends on them.</li></ul><p>The pattern here is subtle but important. None of these are “features” in the traditional roadmap sense. No sales deck leads with resizable table columns. But each one closes a gap where operational error historically entered the system. The product is absorbing risk that used to live in the operator’s head.</p><h3>What DNSSEC and registry incidents keep teaching us</h3><p>DNSSEC deserves special attention because it is the clearest demonstration of the thesis. Cryptographically, DNSSEC works. Operationally, it has been the source of some of the most painful outages in DNS history, and the failure mode is nearly always the same: a signing or rollover operation that was technically permitted but operationally wrong.</p><p>The well-documented incidents follow a script. A key rollover proceeds while stale DS records remain at the parent. Signatures expire because a renewal process silently stalled. An algorithm rollover breaks validation for resolvers that handled the transition strictly. In every case, the system did what it was instructed to do. What was missing was a layer that could say: this change, in this order, at this moment, will break validation, and here is why.</p><p>Registry-level incidents amplify the lesson because the blast radius is not one zone but every zone beneath it. When a TLD has a signing problem, thousands of organizations who did nothing wrong go dark simultaneously. The takeaway for product teams is not “DNSSEC is hard,” which is where most of the discourse stops. The takeaway is that any operation with delayed, distributed failure modes needs product-level guardrails: pre-change validation, dependency awareness, and sequencing enforcement. Documentation is not a guardrail. A runbook that a human must remember to follow is a single point of failure with extra steps.</p><h3>Why UX refreshes are reliability work</h3><p>This is where I want to push back on a common bias. When a DNS provider ships a dashboard refresh, the instinct in technical circles is to shrug: cosmetics, not capability. I think that instinct is wrong, and recent moves across the industry make the case well.</p><p>Recent dashboard refreshes across major DNS platforms are good illustrations. The pattern is consistent: settings that were previously configurable only through the API, things like attack mitigation thresholds, rate limiting, and negative caching behavior, are moving into the interface where they can be seen and audited. Create and edit flows are being redesigned to group interacting settings together. Records management interfaces are gaining advanced filtering, expanded input fields so long values are no longer cut off, and bite-sized in-product explainers for concepts like TTL, delivered at the exact moment an operator needs them.</p><p>Consider what each of those changes actually does to the risk profile. Truncated input fields have caused real incidents: an operator cannot verify a value they cannot see. API-only settings create configuration drift that nobody notices until an attack tests it. Ungrouped settings mean interacting parameters get changed in isolation. In-context education means the person making a TTL decision understands propagation implications at the moment of decision, not after the incident review.</p><p>None of this is cosmetic. It is reliability engineering delivered through interface design. The industry has spent two decades hardening DNS infrastructure against attackers. The harder remaining problem is hardening it against well-intentioned operators at 2 a.m., and that problem is solved in the UX layer.</p><h3>AI-assisted change validation is the next differentiator</h3><p>If guided workflows are the current frontier, I believe AI-assisted change validation is the next one, and it will separate enterprise DNS platforms over the next few years.</p><p>Here is the gap as I see it. Today, even the best DNS products validate syntax and basic semantics. They will stop you from entering a malformed record. What they will not do is reason about intent and blast radius. They cannot tell you that the CNAME you are about to delete is the target of records in three other zones. They cannot tell you that your DS record update, combined with your current TTLs, opens a validation gap of several hours. They cannot tell you that the pattern of changes you are making looks like a migration, and that migrations of this shape historically fail at a specific step.</p><p>These are exactly the judgments a senior DNS engineer makes by instinct, and exactly the judgments that evaporate when that engineer is on vacation. Large language models, grounded in an organization’s actual zone data, query analytics, and change history, are well suited to this class of problem: reviewing a proposed change set, simulating its resolution-path consequences, and flagging risk in plain language before anything propagates.</p><p>The competitive dynamics here are worth stating plainly. Every major DNS platform has comparable protocol coverage. Anycast networks are table stakes. Performance differences are measured in single-digit milliseconds. What is not commoditized is the ability to prevent the outage that would have happened. The platform that can credibly say “our product would have caught that change before it shipped” is selling something no feature matrix captures: fewer incidents, shorter postmortems, and engineers who sleep through the night.</p><h3>The scorecard I would use</h3><p>If I were evaluating enterprise DNS platforms today, I would spend less time on the feature comparison spreadsheet and more time on questions like these. What does the product show me before a risky change is applied? Which settings exist only in the API, invisible to the people auditing configuration? When a DNSSEC operation is about to break validation, does the product know? Where does operational knowledge live: in the product, or in the head of my most senior engineer?</p><p>DNS has been around long enough that the protocol problems are largely solved. The products that win from here will be the ones that treat every incident postmortem as a design requirement, and quietly remove the conditions that made the incident possible. Not more features. Less risk.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9179c9b57fc6" width="1" height="1" alt=""><hr><p><a href="https://blog.stackademic.com/the-best-dns-products-dont-add-features-they-remove-operational-risk-9179c9b57fc6">The Best DNS Products Don’t Add Features. They Remove Operational Risk.</a> was originally published in <a href="https://blog.stackademic.com/">Stackademic</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[When the Signature No Longer Fits the Packet]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://meetcyber.net/when-the-signature-no-longer-fits-the-packet-67ab8440ba62?source=rss-75f0c3cd9f0b------2"><img src="https://cdn-images-1.medium.com/max/2600/1*yLULkMM_HImhRKHrhgYS4Q.png" width="3162"></a></p><p class="medium-feed-snippet">A new quantum executive order quietly sets a deadline that breaks DNSSEC, not just TLS.</p><p class="medium-feed-link"><a href="https://meetcyber.net/when-the-signature-no-longer-fits-the-packet-67ab8440ba62?source=rss-75f0c3cd9f0b------2">Continue reading on MeetCyber »</a></p></div>]]></description>
            <link>https://meetcyber.net/when-the-signature-no-longer-fits-the-packet-67ab8440ba62?source=rss-75f0c3cd9f0b------2</link>
            <guid isPermaLink="false">https://medium.com/p/67ab8440ba62</guid>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[internet-infrastructure]]></category>
            <category><![CDATA[post-quantum-cryptography]]></category>
            <category><![CDATA[dns]]></category>
            <category><![CDATA[cryptography]]></category>
            <dc:creator><![CDATA[Enrique Somoza, D.Sc.]]></dc:creator>
            <pubDate>Wed, 24 Jun 2026 01:04:33 GMT</pubDate>
            <atom:updated>2026-06-24T11:56:17.340Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[The Next AI Revolution Isn’t Coding. It’s Engineering.]]></title>
            <link>https://blog.stackademic.com/the-next-ai-revolution-isnt-coding-it-s-engineering-013601baa06a?source=rss-75f0c3cd9f0b------2</link>
            <guid isPermaLink="false">https://medium.com/p/013601baa06a</guid>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[entrepreneurship]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[business]]></category>
            <dc:creator><![CDATA[Enrique Somoza, D.Sc.]]></dc:creator>
            <pubDate>Mon, 15 Jun 2026 17:57:07 GMT</pubDate>
            <atom:updated>2026-06-19T10:58:56.997Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oso3z98h9hCFsnj52MG8aA@2x.jpeg" /><figcaption>Photo by <a href="https://stevejohnsonart.us/">Steve A Johnson</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Everyone is focused on AI writing code.</p><p>That makes sense. Software is where AI first became visible.</p><p>We have copilots generating functions, agents building applications, and models capable of producing entire prototypes from a simple prompt.</p><p>But the next major disruption may not happen in software at all.</p><p>It may happen in engineering.</p><p>Recently, Jeff Bezos revealed that his startup Prometheus is working toward what it calls an “Artificial General Engineer” — an AI system designed to help engineers create physical products ranging from computers to jet engines.</p><p>Most people see this as another AI announcement.</p><p>I think they’re missing the bigger story.</p><h4>We Already Reduced the Cost of Building Software</h4><p>For decades, the limiting factor in software development was implementation.</p><p>A good idea still required developers, infrastructure, testing, deployment pipelines, and significant investment before anyone could determine whether the idea was worth pursuing.</p><p>AI changed that.</p><p>Today, a single person can go from concept to working prototype in days.</p><p>The cost of validating an idea has collapsed.</p><p>That’s why AI coding tools matter.</p><p>Not because they replace developers.</p><p>Because they dramatically reduce the cost of experimentation.</p><h4>Engineering Has Not Experienced the Same Shift</h4><p>Physical products are different.</p><p>Designing an aircraft component, medical device, semiconductor, or industrial system still requires enormous investments in engineering expertise, simulation, testing, compliance, and manufacturing planning.</p><p>The process is slow.</p><p>Iteration is expensive.</p><p>Validation takes time.</p><p>While software teams have been accelerating, engineering organizations have largely remained constrained by the complexity of the physical world.</p><p>That is the opportunity Prometheus appears to be targeting.</p><h4>The Real Goal Isn’t Better Design</h4><p>The goal isn’t generating a better CAD drawing.</p><p>The goal is compressing the entire engineering workflow.</p><p>Imagine asking:</p><p>“Design a quieter jet engine that improves fuel efficiency by 15% and can be manufactured within a specific cost target.”</p><p>An Artificial General Engineer wouldn’t simply draw something.</p><p>It would:</p><ul><li>Generate design alternatives</li><li>Run simulations</li><li>Evaluate performance trade-offs</li><li>Estimate manufacturing complexity</li><li>Analyze cost implications</li><li>Recommend optimized approaches</li></ul><p>In other words, it would help teams move from idea to validated design much faster.</p><p>The value is not automation.</p><p>The value is acceleration.</p><h4>This Is the Physical World’s Copilot Moment</h4><p>Software developers are already experiencing what happens when AI becomes a force multiplier.</p><p>Engineering may be next.</p><p>The organizations that benefit most won’t necessarily be the ones building the AI.</p><p>They’ll be the companies that use it to innovate faster than competitors.</p><p>Aerospace.</p><p>Manufacturing.</p><p>Automotive.</p><p>Medical devices.</p><p>Semiconductors.</p><p>Any industry where design cycles are measured in months or years could potentially see dramatic reductions in time-to-market.</p><h4>What This Means for Product Leaders</h4><p>Product managers should pay attention.</p><p>The biggest impact of AI may not be that products become smarter.</p><p>It may be that products become easier to create.</p><p>Historically, innovation has often been limited by implementation capacity.</p><p>When implementation becomes cheaper, more ideas get tested.</p><p>More experiments get funded.</p><p>More products reach customers.</p><p>The bottleneck shifts from execution to decision-making.</p><p>That changes the role of leadership.</p><p>The question becomes less:</p><p>“Can we build this?”</p><p>And more:</p><p>“Should we build this?”</p><h4>The Future Is Not Artificial General Intelligence</h4><p>Everyone is debating Artificial General Intelligence.</p><p>The more immediate transformation may come from something else entirely.</p><p>Artificial General Engineers.</p><p>Systems capable of amplifying human expertise across entire disciplines.</p><p>The organizations that learn how to combine human judgment with machine-driven engineering may gain an advantage that is difficult to replicate.</p><p>We saw it happen in software.</p><p>Now we’re about to find out what happens when it reaches the physical world.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=013601baa06a" width="1" height="1" alt=""><hr><p><a href="https://blog.stackademic.com/the-next-ai-revolution-isnt-coding-it-s-engineering-013601baa06a">The Next AI Revolution Isn’t Coding. It’s Engineering.</a> was originally published in <a href="https://blog.stackademic.com/">Stackademic</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>