Sitemap
Rob Chandhok

Rob Chandhok

Recovering CTO. Retired.

Thing Neutrality

4 min readSep 25, 2014

--

Balkanization: to break up (as a region or group) into smaller and often hostile units…The term is also used to describe other forms of disintegration, including, for instance, the subdivision of the Internet into separate enclaves.

We are at a fascinating turning point in the evolution of the Internet. So much of the tech communities’ focus has been on the app ecosystem and the cloud that we have left the messy edges of the Internet to grow up on their own. Different protocols, different radios, different models for security and privacy and trust.

That design autonomy at the edges is about to be disrupted by the the phenomena we are currently calling the “Internet of Things”. The term started out referring to new kinds devices that we were connecting to the Internet but has morphed into a term describing devices, sensing data, and even data in the large.

What’s the problem?

Some folks don't see a problem at all. The Internet is the Internet, after all. Even the phrase “the Internet of Everything” deconstructs to “the Internet” if you really get down to it — the very meaning of INTERnet is that internetworking between heterogeneous devices.

The core of the Internet kept true to a strongly layered design principle. The protocol layers (IP and upwards) were very separate from the physical layer protocols. I remember when we (at CMU) first got DEC bridges that actually filtered traffic on our Ethernet cables, segregating traffic to increase local capacity. This “new” feature and a host of others all happened without breaking or changing new TCP/IP layers like SMTP or FTP (HTTP wasn't around yet).

Get Rob Chandhok’s stories in your inbox

Join Medium for free to get updates from this writer.

The problem is that the proximal side of connectivity hasn’t kept true to that design principle. Not because the designers and innovators were lazy, or less smart, but because the initial constraints on the problem (radios, power, chip size, etc.) were so overwhelming that it often required building intertwined systems to satisfy the basic connectivity requirements. These systems were incompatible, not IP based, and not interoperable. So we ended up with fully different solutions in unlicensed bands and even in the same band (the 802.15.4 families, for example).

The end result is a system that is counter to those original design principles of the Internet. We don't talk about a cloud server as “ethernet” or “WiFi”. But we talk about “Zigbee” switches and DECT phones and other devices by their radio connectivity. And the radios in these devices are deeply intertwined with the protocols. This would never be successful on the core of the Internet. HTTP would have been a failure if Tim Berners-Lee had said he had a great new hypertext protocol but it only worked on token-ring!

A new approach

The time has come to bring back those original principles with a recognition that there are new challenges as well. Let’s build systems where the radio folks can innovate without interfering with the networking protocol folks, and vice-versa. Let’s keep the interoperability of the Internet without forgetting that the Internet may need to grow and change topologies at the edges to accommodate new kinds of nodes, new use cases, and different usage requirements.

We need to stop allowing the excuse of “it’s too hard to do unless we optimize completely”. In my opinion, that is a lazy approach to the problem. We can afford to take the time to rethink now that we have far better radio and compute technology on which to base this.

This means that the standards bodies responsible for device or peer to peer radios need to stop walking their way up the stack. We shouldn't have application layer protocols that are specific to a kind of wire or radio. We shouldn't have radio standards bodies dictating how software developers should implement services.

We should no longer accept that new application layer services require firmware upgrades.

Some would argue that we just can't make IP work at low power, and that without squeezing every last literal bit of efficiency we can't move forward. That is an example of early optimization. The folks doing Thread appear to have found a balance with efficient radios that implement IP correctly. That means we can layer other platforms and solutions right on top, like Alljoyn or MQTT. That’s the right system partitioning.

If the Internet of Things industry fails to recognize that the layering that drove innovation and robustness in the Internet has to come to devices then they will simply be left behind by the Internet software posse. Dividing up the pie by physical connectivity type is not what grows ecosystems, so as the Internet of Things grows, the connectivity for things must be neutral and open. Software eats the world, as Marc Andreessen says, and in this case the edges of the Internet are far too tasty to ignore.

--

--