Re: comments on draft-ietf-tcpsat-stand-mech-04.txt

[Vern Paxson <[e-mail]>]

> [Aaron wrote:]
>     I don't think that because it is only good in some 
>     satellite scenarios that it now becomes a research issue.

Acking every packet isn't only good for satellite scenarios, it helps
performance in general.

> 1) Is it legal with respect to RFC 793 (and 1122, and...)?
> ...
> (1) - yes, but discouraged in spirit by RFC-1122's *should*
>       clauses regarding delayed acks.

Depends on how you interpret the fact that there's no minimum value
required for the delay timer.  (But I will admit that interpreting
zero as being just fine is a tad squirrely. :-)

> 2) Are we violating fairness/Are we sure that it won't 
>    degrade performance for other connections on shared paths?
> ...
> (2) - Is it OK for some connections to be more aggressive than
>       others? If so, how does a client *know* that it traversing
>       a long delay path instead of a terrestrial path? Where is
>       the proof that this is a safe thing to allow for general use?

I should've been more clear - we don't just do this for satellite paths,
we allow it for TCPs in general.

> Finally, since "byte-counting" is the moral equivalent of acking
> every packet

No, it has a major difference.  With byte counting, the window advances
in bursts, and gets *burstier* when there's loss along the return path.
(Byte counting is also clearly not consistent with existing RFCs.  Note
also that when most researchers do TCP simulations, they simulate
ack-every-packet; it's already a de facto element of that part of the
research community.)

> I wasn't in Munich (although my slides were) and get feedback 
> telling me that both these possibilities met with some fairly 
> hostile response. That was the last time the subject was broached 
> in this forum - have things changed?

What has changed for me was Van Jacobson strongly arguing for
ack-every-packet.  I'm not fully convinced that it's sufficiently
conservative (and a win), but am fairly convinced.  So I wanted to
float it by the list as part of my comments to see what others think.
I'm also asking the Transport directorate for their views, to see what
sort of consensus we have.

> If the satellite subnetworks were guaranteed to be separated
> from the Internet backbone (say, via a proxy or splitting gateway), 

Again, I'm not viewing ack-every in that context.

Finally: I don't have my heart set on this!  But am quite interested in
exploring it as a standard mechanism, as it strikes me as (1) significantly
improving performance, especially for high latency paths, (2) being
basically already legal, (3) having the support of some eminent congestion
control researchers.

		Vern