[e2e] Open the floodgate

[Cannara]

Thanks Craig (since I'm in the list, you needn't copy my address),

To again reduce the traffic, I'll merge your other msg...

    Mistaking an error-induced loss for congestion loss hurts, but
    mistaking congestion loss for error-induced loss hurts much much
    worse.  (Even if we're talking congestion loss on a link with lots
    of errors).

Very true.  That's why it's important that the network understand its own
congestion and, at least, inform layers above of what's going on.  No one
disputes the need for congestion control, because it's done at all layers in
various parts of most any network.  The issue is that work stopped after TCP
was saddled with the stopgap job of managing IP congestion in a partially-
blind way.  VJ's remark should have been impetus for continued development,
but wasn't, for various reasons that various folks around at that era
understand.  

Today, we have various transports doing their own congestion management in
their own ways, and some still not getting benefit.  There are people
revisiting the need and suggesting better network architectures.  Great!

On the XCP meeting and Greg C., etc., I don't know him and I'll believe he was
difficult.  But, we still have suboptimal network design, insofar as we depend
on TCP from the '80s and a glacial IETF process -- all this while now having
complete web servers on a chip inside an RJ45 jack!  So maybe his ideas for
SiProts were something to consider, even if they weren't right on target? 
Personalities and provincial behavior shouldn't define what we have these many
years later.  Business desires work around disorganized designs and procedures
anyway, which leads to even more fractured results.  That's my concern.

Alex

Craig Partridge wrote:
> 
> In message <4086B2FB.58ED16A6 at attglobal.net>, Cannara writes:
> 
> >"The High Priest of XCP came to one of our meetings to describe his
> >religion.  The discussion was polite, but we were not convinced."?
> 
> You've obviously never talked with Greg Chesson when he's passionate
> about something :-).
> 
> >Is the "we" the royal one?  Why the references: "High Priest", or "his
> >religion"?  Sounds like religion on both sides maintaining suboptimal network
> >design for the masses.  Which is, of course, what has happened.
> 
> As I recall the discussion (and yes, "we" is right as there were about
> 15 to 20 folks in the room), we struggled to find a common ground to talk.
> Greg wanted to talk about how all protocol problems could easily be reduced to
> a silicon processing engine if we were willing to rethink the problem, and
> most of the E2E folks in the room said "so how do you plan to rethink
> congestion control in the network and memory management in the end system?"
> and Greg didn't want to talk about those issues.  Very frustrating all
> around.
> 
> Craig
> 
> PS: In years since, Tim Strayer, who did a lot of the legwork on XTP, and
> I have ended up working together and we periodically chat about XTP and
> have a good time doing so.