Re: [arch-d] FYI: closure of the IAB Stack Evolution program

[John Leslie <[e-mail]>]

Henning Schulzrinne <[e-mail]> wrote:
> 
> Given that the economic incentives are against providing such services,
> and competition in most jurisdictions is weak or largely non-existent
> (US) and that competition for user-invisible features is difficult,
> you are asking for regulatory intervention. (Nothing wrong with that -
> you could consider this a version of a market failure.)

   It _is_ a market failure, of course; but hardly a failure of a _free_
market. There is no such thing in communication services.

   The proper question, of course isn't whether it's "wrong," but rather
whether "regulatory intervention" is likely to improve things.
 
> The additional problem is that you need near-100% compliance to allow
> systems designers to rely on these capabilities.

   ROTFL.

> Thus, unless almost every ISP allows non-UDP/TCP transport layers,
> nobody is going to be building such layers.

   The classic statement, of course, is that <evil-action> may be
considered "damage" and we "route around it."

   (I thought that was well-known; and we agreed upon it. But I may
be wrong!)

   There is already quite an infrastructure of "virtual private
networks" to bypass censorship choke-points. The same solution can
be set up to bypass other internet choke-points; and I wouldn't be
surprised to see these become popular...

   ... when individual users feel the pain.

   Alas, too many of _us_ expect to replace sub-optimal government
policies with good government policies. (I was guilty of that
myself fifty years ago.)

   Others of us place our faith in government incompetence (which
faith seems well-founded) and think we can use international
organizations such as IETF.org to trump their incompetence. YMMV...

   My attitude is that our salvation lies in numbers: that traffic
_will_ increase, and overheads which were intolerable thirty years
ago will pale into insignificance and no government will be able
to search for them.

   Thus "tunneling" in the generic sense will become our favored
tool. But I'm not wasting my time trying to define the perfect
tunneling tool; least of all chasing the "perfect" tunneling
architecture.

   (Right now, alas, I'm looking for a way to "tunnel" around 30%
packet loss, which is this week's "damage" being inflicted upon
New Hampshire by the incompetence of our too-few home Internet
services.)

> As a matter of terminology, we also seem to be re-iterating the old
> discussion that layers are no longer quite as simple as they used to
> be in the 1980s, where the function of a layer can now be broken up
> into multiple sub-layers, shims or whatever else you'd want to call
> them.

   "shims" is a good name.

   I'd love to facilitate that discussion. I'm not sure I can, since
the backgrounds of our participants vary so much.

> As an older example, you could argue that RTP is a (specialized)
> transport protocol, used with UDP or TCP, for timing-sensitive
> applications.

   ;^)

--
John Leslie <[e-mail]>