Re: Call for Papers: IAB Workshop on Stack Evolution in a Middlebox Internet (SEMI)

[Joe Touch <[e-mail]>]

On 12/11/2014 6:03 AM, Phillip Hallam-Baker wrote:
> On Wed, Dec 10, 2014 at 8:26 PM, Joe Touch <[e-mail]
> <mailto:[e-mail]>> wrote:
> 
...
>     > The Internet architecture to date has been what survived a
>     > Darwinian process.
> 
>     First, it's not Darwinian so much as mutation caused by high-energy
>     radiation, and no, it's not clear to me that "architecture" is
>     surviving. Sometimes the result is just glowing goo.
> 
> What is surprising is that it is possible to describe what has survived
> in a remarkably clean fashion with almost no recourse to special casing
> except on the issue of syntax.

If that were true, we would have such a description in hand, and the IAB
wouldn't need to run a workshop to figure out how middleboxes fit into it.

> The glowing goo rarely survives long. My problem is that things get
> described as glowing goo for no other reason than that we didn't think
> of them.
> 
> Are we really taking the high ground in the architecture discussion here
> or have we merely staked a position in an editors war?
> 
> The argument seems to be that we define the architecture, therefore
> anything that isn't the architecture is wrong is an abomination and
> since the only things that don't fit the architecture are abominations,
> this proves we don't need to reconsider the architecture.
> 
> It is a hermetically sealed system of thought. 

Let me clarify.

An architecture describes a system of components and their interactions
in an internally consistent way, which can be used to reason about behavior.

Architectures aren't fixed things; they can evolve, e.g., you can go
from one stable architecture to another, more capable one by deliberate
extension.

It's much harder to just allow unrestricted "innovation" (i.e., anarchy)
and then try to back-calculate a valid architecture that includes those
innovations. Sometimes it's possible, but most of the time the
innovation as deployed wasn't designed to be consistent with a new
architecture. That doesn't mean you can't do the extension - it just
means you can't do an unrestricted extension.

If we had an IAB, they'd be dealing with this evolution as we go, rather
than trying to back-fit it after extensions are widely deployed without
constraint.

> The real heart of the Internet isn't the narrow waist or even the
> end-to-end principle. It is that the Internet is designed to support
> novel functionality. The narrow waist is probably an essential
> consequence of that goal. 

The real heart of the Internet is an architecture. The E2E principle and
narrow waist are consequences of that architecture. If you put either
one ahead of that, you just have a bumper sticker with no utility.

Joe