Re: [Tsvwg] Input for draft-paxson-tcp-rto-00.txt

[Reiner Ludwig <[e-mail]>]

At 00:42 25.11.99 , Vern Paxson wrote:
> > Why would we (TSVWG) want to do that when we know that the current TCP
> > retransmission timer as documented by draft-paxson-tcp-rto-00.txt has some
> > major problems, e.g., those discussed in our paper?
>
>Same reason that RFC 2581 (Proposed Standard) doesn't address some of the
>fast recovery problems that, e.g., RFC 2582 (Experimental) does.  The scope
>of draft-paxson-tcp-rto-00.txt is to document the de facto standard, not
>to refine it.

OK. I assume that this is the consensus of the WG.

In that case, however, I'm confused about the purpose of a Proposed 
Standard. I thought that when the IETF publishes an RFC as PS then it says: 
"To the best of our knowledge, this is how it should (has to) be done."

I thought that things like documenting common practice *and* the problems 
with that were published as an Informational RFC.

The case with RFC 2581 was different. Even if there still were/are some 
corner case improvements to TCP's congestion control, there was no major 
problem with it, and the IETF could rightfully say: "To the best of our 
knowledge, this is how it should (has to) be done."

For TCP's retransmission timer this is not the case. It has major problems. 
I agree, though, that it's not a threat to the network because the 
combination of all problems results in a (too) conservative retransmission 
timer. But is that enough to justify a Proposed Standard?

///Reiner