[Chapter-delegates] Unfortunately, the Internet Impact Toolkit is unsound

[Andrew Sullivan]

Hi Dave,

Thanks for your remarks.

On Wed, Sep 09, 2020 at 12:29:49PM -0400, Dave Burstein wrote:

>Columbia Professor Eli Noam, perhaps the Internet's most respected public
>voice, pointed out that there was no reason what's valuable about the
>Internet required a single system. So long as there was robust
>interconnection, there could be many "Internets" with different internal
>systems.

That _is_ how the Internet works.  It's the reason your wifi and your phone can both be connected to the Internet at once.  But they use a common underlying protocol to do it -- a common protocol that permits interconnection in an end to end fashion.  It's certainly true that it need not have been IP.  But it worked out that "robust interconnection" required an end to end approach.

If what you mean by "robust interconnection" is actually "good gateways between otherwise independent networks, with controls at the network borders", it's certainly true that you could build a network that way.  Indeed, the world did: it's the phone system.  And early work predating the Internet actually also worked that way: it's the CATANET.  Depending on which features of 5G one selects, you could get the CATANET back out of it too.  But that wouldn't be "many internets".  It would be a CATANET.

>There is absolutely no reason the future Internet requires "a Common
>Protocol," "a Single Distributed Routing System," or "Common Global
>Identifiers"* I think those are good ideas, but can respect those who
>disagree. *

If you want the Internet, we claim that these are actually necessary conditions.  That isn't to say that you need these for _any network_.  You just need it to have anything we can recognize as the Internet.

>Bob Kahn, a founder of ISOC, supports DOA/DONA, a "digital object
>architecture."

Sure.  And there's nothing wrong with that or incompatible with the Internet. 

>In another example, Verizon, backed by Deutsche Telekom, Telefonica, the
>Koreans, and the Chinese want "deterministic networking" designed to
>guarantee performance.

_That_ system, on the other hand, doesn't appear to be the Internet, but actually a specialist network that lives along side the Internet and which, if it interconnects, is likely to end up using the Internet way to interconnect.

>be effective, we should clearly take positions that benefit Internet users
>around the world.  We must be very careful not to take positions that
>protect U.S. giant companies.

I don't believe anything in the toolkit released today inherently protects giant companies, USian or otherwise, and I can certainly think of giant companies for whose systems architecture the toolkit provides pretty serious critical tools.  Perhaps you can suggest where you think we've missed the mark, because I'm having a hard time understanding your critique in the general terms in which you seem to be making it.

Regards,

A

-- 
Andrew Sullivan
President & CEO, Internet Society
sullivan at isoc.org
+1 416 731 1261