Re: [tcpm] timestamp options (was Re: New Version Notification for draft-touch-tcpm-tcp-edo-01.txt)

[Mark Allman <[e-mail]>]

> I think it was Mark Allman who questioned a while ago whether the
> benefits of the timestamp option are worth spending 10 bytes in every
> packet. (But I can't find the email now.) 

Just to be clear ...

  - The RTTM scheme for getting a better RTO seems generally dubious to
    me (based on an old paper I wrote with Vern).  It does not seem
    worthwhile to spend 10 bytes / packet to get more RTT samples per
    RTT from the perspective of getting a better RTO (because we don't).

  - However, if you're sending fast then you still have a sequence wrap
    issue.  The only standard way that I am aware of that we deal with
    that issue is timestamps in every packet.  So, for this I certainly
    think there is value in having the TS option there.

That said, there is ample opportunity to turn TS on in the midst of a
connection before we need it for PAWS, but allowing the vast majority of
the connections to save the bytes.

allman