[Tsvwg] Re: [e2e] What's the benefit of out-of-order processi

[David Borman]

> From: John Wroclawski <jtw at lcs.mit.edu>
> Subject: Re: [Tsvwg] Re: [e2e] What's the benefit of out-of-order processi
> Date: Mon, 17 Sep 2001 22:31:55 -0400
>
> At 6:36 PM -0700 9/17/01, Jacob Heitz wrote:
> >The urgent pointer does not point at a range, only at a single
> >byte. It can only deliver a single byte of out-of-band data.
> >
>
> Actually, TCP urgent delivers an out of band _notification_ of the 
> presence of urgent _in band_ data.
> 
> (The idea being that on receiving the notification the app would know 
> to chug through the data stream in whatever way it sees fit until it 
> gets to the 'urgent' part.)

A minor nit.  All the data up to, but not including, the urgent pointer
is the "urgent data".  The urgent pointer points to the first octet
following the urgent data.  But in most cases, what the application
is really interested in is the data after the urgent pointer, sometimes
referred to as the "interesting" data.

The other thing to keep in mind is that the urgent pointer can move,
and is cumulative.  Thus, the urgent pointer might point at "X",
and can then be moved to "X+n" before "X" is processed, or even
sent!  This is how TCP deals with urgent pointers that have more
than a 64K offset.

But I digress.  John is right, the TCP urgent pointer is basicly
out-of-band *notification* of interesting in-band data (and what
"interesting" means is determined by the application).

		-David Borman, dab at windriver.com