Re: [tsvwg] TCP Prague's RFC 5033 guidelines status?

["Holland, Jake" <[e-mail]>]

On 2019-11-11, 14:12, "Wesley Eddy" <[e-mail]> wrote:
> I believe you're referring to the congestion response requirements in 
> the L4S ID draft.  I think those are a start to requiring an end-host 
> algorithm to have good properties with regard to 5033 guidelines, rather 
> than requiring some bad behavior (e.g. the first one is "MUST coexist 
> safely with Reno" which speaks right to 5033 guideline 1 "Impact on 
> Standard TCP, SCTP, and DCCP").  So, I'm not sure why those are a 
> concern (i.e. which L4S ID congestion response requirements do we have 
> 5033 concerns with?).

I don't want to get too bogged down on venue issues, and the point that
TCP Prague is out of scope for tsvwg is fair.

Maybe the issue that's appropriate in the tsv tracker would be something
more like "a reasonably complete candidate definition of at least one
proposed congestion control suitable for deployment on the internet that
makes use of the proposal", and details like asking for an RFC 5033-related
section could be handled separately in reviewing those docs?  After all,
I'm not sure there's much urgency on starting the experiment if there's not
a candidate endpoint system defined...

(This is exacerbated by the fact that DCTCP didn't do this, since it had
disclaimers like "only in controlled environments" and "documents what's
currently deployed", but now that it's proposed for the internet, it seems
to me the evaluation should be written somewhere in order to run this
experiment.)

Anyway, the motivation for the request was because it's hard to go through
the 5033 requirements and check them to decide how well they've been covered.
For example: the difficult environments considered (2) and the range of
environments investigated (3), performance with misbehaving nodes and
outside attackers (6), and responses to sudden or transient events (7).

I was thinking there have been a number of tests along these lines done as
sort of random spot checks, but it's hard to go through the list of tests
looking at responses to sudden or transient events to figure out whether a
case you're worried about is covered, and likewise with the others.

So I was asking for the authors to apply their uniquely wide knowledge of
the tests conducted toward categorizing the references in the way suggested
by RFC 5033, maybe in an appendix or a side doc, so they could be more
readily evaluated.

I agree some of the 5033 requirements are likely covered to some extent
already (#1 I agree seems reasonably well examined, as it's a consistent
theme in many of the reports, though it would still be useful to have a
list of references to areas where it's directly addressed), but I'm less
sure of several of the others.

Even a list of what 5033 requirements are covered and not covered by the
L4S requirements to set expectations on the ones that TCP Prague or other
L4S-compatible congestion controls will have to explicitly address would
be worthwhile, IMO.

Thanks and regards,
Jake