Re: [Tsvwg] Re: [tcpm] Revision of draft-larsen-tsvwg-port-randomization

[Joe Touch <[e-mail]>]

Murari Sridharan wrote:
> Agree on the NAT part.
> 
> However whether or not its negotiated during SYN exchange would
> depend
> on how it gets specified. For instance (again I really am thinking
> real-time and aloud here not fully thought thru) we could do something
> like the following. On the SYN packet, the sender picks a port
> (unscaled)and tries to negotiate a scaling option, (NATs that do packet
> inspection are free to strip this option out) and the destination picks
> a port and may respond to the scaling. If both sides agreed, then the
> scaled ports are used for the 4-tupe, if not the unscaled version is
> used. The only caveat is that while a connection attempt is in progress
> to a destination, we cannot be trying the same local port, since till
> the 3-way handshake completes you wont know whether the scaled or the
> unscaled version is going to be used.

It can't get negotiated during the SYN; NATs wouldn't track that, and
will break the connection if it 'moves' to another port.

Even if it is in the SYN, it won't allow two hosts to share a port
through a NAT.

It's worth looking at the alternatives which are discussed in Mark's ID
and mine, too.

Joe

-- 
----------------------------------------------------------------------
Joe Touch                Sr. Network Engineer, USAF TSAT Space Segment
               Postel Center Director & Research Assoc. Prof., USC/ISI

_______________________________________________
tcpm mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/tcpm