(Tom Herbert, SiPanda CTO, April 28, 2021)
The Internet is one of humanity's greatest achievements. Think about it, at our fingertips or voice command we have instant access to the vast domain of scientific knowledge, art, entertainment, and real-time world-wide communications. The Internet is much like the Emerald City in the Wizard of Oz, a marvelous and seemingly magical place. But the Internet doesn’t run on magic: there is an immense underlying infrastructure making it happen. Similar to the grand revelation in The Wizard of Oz, when the curtain is lifted on the core infrastructure of the Internet allowing us to see behind the scenes, not everything is as we hoped. The Internet has become ossified, or “ozzified” if you will, and the aspirations of the original architects in the 1960's and 1970’s for the Internet to be a platform of unlimited innovation have been mostly forgotten. One might argue that the Internet is a victim of its own success and decisions made by the “wizards” behind the scenes were prudent to expedite a ubiquitous world-wide network; perhaps, but nonetheless the continued evolution and innovation of the Internet is hampered by its ossification.
To understand why the Internet has become ossified, one needs to understand how the Internet works. In the simplest model, standards organizations, like the Internet Engineering Task Force (IETF), define standard, interoperable protocols which are then implemented in applications, end hosts, and routers charged with processing and moving packets and data through the Internet. On paper, as long as everyone implements the standards as specified, all of this works and the result is an interoperable and extensible world-wide network. Reality is much different. In order to make feasible implementation, particularly in hardware that needs to move packets at high speed, some networking vendors have made trade-offs that limit both the protocols and features that they support. In the most degenerative case, vendors may only support the most basic protocol formats, for instance plain TCP or UDP over IPv4 or IPv6. Not all vendors make these tradeoffs, but if a small but significant portion of them do, that drives down the Least Common Denominator of support, effectively limiting the potential of everyone. The result is that it is very difficult to introduce protocols or features on the Internet and to innovate, hence we say the Internet has become “ossified”.
An exemplar of a useful feature commonly sacrificed for the sake of expediency is Type Value Lists or TLVs. TLVs are a common networking protocol construct composed of a list of data elements each where each element contains variable length typed data. TLVs are very expressive and make protocols extensible, as IPv6 Hop-by-Hop Options make IPv6 extensible, but they are generally considered hard to process, especially in hardware, because of their variable length and combinatorial ordering. Since conventional wisdom holds that one cannot simultaneously achieve flexibility and high performance, many router implementations have simply punted on TLV support and may unilaterally drop packets containing TLVs.
At SiPanda, we don’t subscribe to the charlatan Wizard’s facade of brushing off the hard problems with smoke and mirrors or rationalizations. To the contrary, we embrace the challenge to solve the hard underlying problems that are holding the Internet back. For instance, we believe efficient TLV processing in hardware is not an unsolvable problem but is facilitated with the right application of technological innovations and some practical constraints. To be certain, the road to undoing the Ozzification of the Internet will not be easy (much like Dorothy’s trip up the Yellow Brick Road), but if we don’t start the journey we’ll never get to where we want to be.
Comments