Scheffenegger, Richard <[email protected]> wrote:
> [Michael Welzl <[email protected]> wrote:]
> 
>> Do you know about any other disadvantages?

   While I believe discussion of disadvantages would help, I see a more
fundamental question:

   "What modifications/extensions to RFC 3168 are appropriate to achieve
the benefits being discussed?"

   (Of course, RFCs 4301 and 6040 discuss handling of ECN in tunnels;
this much extension is hopefully a given.)

> I think the biggest issue is the interaction between ECN and non-ECN
> flows (and packets;

   I'm not quite sure what Richard means here -- there's an obvious
question about one starving the other (and I do think that's an issue,
but not quite a "disadvantage").

> for TCP, only data segments are specified to optionally be ECN-enabled,
> not pure ACKs, and other control segments (some are allowed as
> experimental, i.e. SYN/ACK).

   This is an obvious area for improvement; but I really can't see it as
a "disadvantage" of ECN.

> If ECN would have been seen as universally good (back in the day),
> it is odd that one L4 flow (TCP session) is constituted from both
> ECN- and non-ECN segments during it's lifetime.

   I don't see that as "odd" -- just "suboptimal".

> The natural conclusion would be to assume that ECN on pure ACK, FIN, RST
> is detrimental in some way (personally I tend to disagree; there may not
> be any significant benefit, but it wouldn't break anything either -
> just the opposite, pure ACKs would serve as probes into the congestion
> state of the return path, once the data direction within the TCP flow
> reverses).

   Again, an obvious area for improvement...

> But these points are transport specific; 

   Exactly! It makes very little sense to blame "ECN" for these issues.

> However, guidance to transport protocol designers on the use of ECN
> with control segments etc seems like it could be in scope for this
> document?

   I don't agree.

   The Abstract says:
" 
" This document describes the potential benefits to applications when
" they enable Explicit Congestion Notification (ECN).

which implies to me that it should discuss benefits regardless of the
particular transport protocol (and I think it does).

   The document describes setting the ECT(0) or ECT(1) codepoint in the
IP header and how routers may react to that. Not until Section 4 does
it mention particular transports; and IMHO those are listed strictly
as examples.

   I believe the document intentionally avoids questions about the
design of transport protocols; and I think it should continue to avoid
those.

   But the document _does_ mention several times the desirability of
ECN-marking packets which would not otherwise be dropped.

   I quite agree that is desirable; but it is a fundamental change to
RFC 3168; and I think we need to justify it considerably more than
this document does. This will not be a trivial change to deploy.

   It will indeed require deployment in routers and transport stacks,
but the question is quite independent of the transport _protocols_ --
and I strongly suggest we don't confuse the two.

--
John Leslie <[email protected]>

_______________________________________________
aqm mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/aqm

Reply via email to