Re: [Tsvwg] ECN & PMTU

[Kostas Pentikousis <[e-mail]>]

On Tue, 9 Apr 2002, Jacob Heitz wrote:

[...]
  | Source quenching was always controversial.
  | It is considered higher impact, because it adds
  | an extra packet to the network. Setting a CE
  | bit adds nothing that was not there before anyway,
  | so is considered to be safer, or lower impact.
  |
  | Yes, it's slower than having the congested router
  | send an ICMP.

I'm not sure that's really true. In fact, I've heard several times that Source
Quench (SQ) via ICMP is more expensive for current generation routers than
setting a bit. I remember reading that any packets needing "special" treatment
are placed in a separate (lower priority) queue and are dealt with when the
router load drops (a bit). Setting a bit is not considered "special" treatment
but creating a packet (setting source and destination addresses, etc.) is. If
this is true, and ECN was spec'ed so that the router would send an SQ ICMP
packet back to the sender instead of setting a bit (RFC 3168), then in most
cases the congestion notification would actually arrive later than an RTT.

Just my $0.02,

Kostas
______________________________________________________________________
Kostas Pentikousis @ CS Stony Brook @ http://www.cs.sunysb.edu/~kostas