Re: [tcpm] TCP tuning

[Mark Allman <[e-mail]>]

[These are re-sends from long ago... they never made it to the list.
 But, that seems to have been worked out, now.  --allman]


> We do need to be careful that whatever we do doesn't overwhelm these
> assumptions and change the game, however.

Agreed.

> Regarding init cwnds, I'd like to see a reasonable level of analysis
> of the impact. 

Yep.  I wasn't trying to say we'd just make a WAG.  Certainly we'd want
to see an assessment.  That said, personally, I have a lower bar for
letting this sort of thing through than perhaps I have had in the past.

(I think the initial window argument about most transfers being small is
compelling.  But, I wrote a paper (**not IETF proposal**) with some
colleagues a couple of years ago [*,**] that wondered about just getting
rid of slow start.  There are some meager simulations, but also some
simple arguments as to why this might not be an awful approach.)

> It's not enough to drive discussions with "we have to RIGHT NOW!" to
> satisfy a (often underinformed) perception of need.

Absolutely.

allman


[*] Dan Liu, Mark Allman, Shudong Jin, Limin Wang.  Congestion Control
    Without a Startup Phase.  Protocols for Fast, Long Distance Networks
    (PFLDnet) Workshop, February 2007.
    http://www.icir.org/mallman/papers/jumpstart-pfldnet07.pdf

[**] The related story is that this paper was the start of my undo-ing
     as TCPM co-chair as Lars' comment after seeing the paper at PFLDnet
     was something like "can we really have such a whack job chairing
     TCPM?".  [***]

[***] OK, OK.  While I recall Lars making this comment, I have no
      evidence that it was my downfall at all.  Especially since I
      voluntarily resigned.  Um, I think I did, anyway. :)