[e2e] How prevalent is Timestamp option and PAWS?

[Mark Allman]

Anil-

> I wonder how prevalent is the use of PAWS (Protect Against Wrapped
> Sequence Number) in modern TCP implementations ? As per RFC 1323
> (section 4, page no. 17) any TCP using timestamp can optionally
> implement this.  But, has it really been implemented in modern
> TCPs? Any statistics?
> 
> Also, any statistics on No. of hosts using timestamp option as
> specified in RFC 1323

There is some data on the subject in the following paper:

    Mark Allman. A Web Server's View of the Transport Layer. ACM
    Computer Communication Review, 30(5), October 2000.
    http://roland.grc.nasa.gov/~mallman/papers/webobs-ccr.ps

The data is a bit old at this point (from 1998 -- early 2000), but
might offer some clues.  I am not aware of any other measurements.

(I really should dig into the pile of data I have since early 2000
and get some new numbers.  A job for my copious spare time...)

The measurements likely indicate what hosts do by default.  As far
as I know, all major operating systems now have the ability to do
the RFC1323 extensions (but, most require some tweaks to use these
by default).

I hope that helps.

allman


--
Mark Allman -- BBN/NASA GRC -- http://roland.grc.nasa.gov/~mallman/