In message <20050302210717.GB3277@s102-n054.tele2.cz>, Jaromir Dolecek writes
>On Thu, Mar 03, 2005 at 07:45:23AM +1100, Daniel Carosone wrote:
>> On Wed, Mar 02, 2005 at 09:42:01PM +0100, Jaromir Dolecek wrote:

[...]

>> We also coalesce mbufs (differently, at the moment) in tcp fragment
>> reassembly.
>
>Do you mean ip_reass() ?
>
>AFAICS there is no mbuf coalescing happening there besides m_cat().
>And m_cat() is exactly TRT there.

I assume Dan means sys/netinet/tcp_input.c:tcp_reass().  It also
brings sys/kern/uipc_socket2.c:sbcompress() to mind --- at least, to
my mind --- and mabye the combination of m_cat() and later sbcompess()
is what Dan was thinking of, too.