Subject: Re: ip reassembly time exceeded?
To: Olaf Seibert <rhialto@polder.ubc.kun.nl>
From: Wiker, Raymond <etorwi@eto.ericsson.se>
List: current-users
Date: 01/30/1997 18:00:03
Olaf Seibert writes:
 > My guess is that M$'s reasoning is about as follows. (Just a guess of
 > course.) It receives a bunch of fragments and reassembles them. It gets
 > a datagram that is smaller than the 576 bytes and thinks "well, that's
 > so small, why would anybody want to fragment that? Let's wait for some
 > more..."

	I don't think so, because the fragemnts of an IP packet
contains sufficient information to tell that all fragments of a packet
has been received.

	Another possibility: MS's IP stack (or applications using IP)
may use Path MTU discovery to find the optimum packet size. If the
Path MTU discovery algorithm sets 576 octets as the lowest acceptable
Path MTU, the process fails. Alternatively, Path MTU discovery returns
a value less than 576, and the IP stack or the application decides
that there is a buggy router/host out there, and refuses to talk to
it.

	//Raymond.