Re: [tcpm] TCP zero window timeout?

On Sat, Aug 26, 2006 at 04:05:45PM -0700, MURALI BASHYAM wrote:
> 
> Fernando
> 
> I collaborated with mahesh on this, so let me try
> making the case for it a little better.
> 
> The problem was found on a TCP proxy, which does not
> have any applicaton awareness. In fact application
> awareness is not a goal in that environment. Hence the
> TCP level solution.

(As just a participant)

Why get TCP involved here?

If a proxy is running out of memory and it has connections that are
largely unused, it should close them.  The proxy is at the exact right
spot to keep track of the connections and their usage.  It can tell
which connections are wasting its resources and it knows what it
considers bad behavior of a connection.  When proxy resources get
scarce, close the connections that are behaving the worst - have been
holding the same proxy resources a long time, for example.

None of that requires application-awareness or TCP's complicity.

And even if it did, TCP enables many applications.  To a first
approximation, increasing TCP's complexity increases its fragility.
Changes to TCP should benefit more than one application before they're
standardized, IMHO.

-- 
Ted Faber
http://www.isi.edu/~faber           PGP: http://www.isi.edu/~faber/pubkeys.asc
Unexpected attachment on this mail? See http://www.isi.edu/~faber/FAQ.html#SIG