Subject: Re: ip reassembly time exceeded?
To: None <current-users@NetBSD.ORG>
From: Christoph Badura <email@example.com>
Date: 01/30/1997 03:56:00
jonathan@DSG.Stanford.EDU (Jonathan Stone) writes;
>>Perry Metzger writes:
>>You are required to be able to REASSEMBLE a 576 byte packet. You are
>>NOT required to be able to transport one, although it is recommended
>>that you be able to do so.
>Perry is of course correct. See rfc1122 for authoritative details.
>The 576-byte MTU requirement is an oft-repeated canard for IPv4. So
>much for Microsoft Internet engineering.
Except, that in this case the router at Jukka's end was fragmenting
552 byte IP packets into smaller ones and it is entirely unclear why
Microsoft's mailservers can't reassemble the packets. For all we know
their firewalls could drop the some of the fragments.
>However, I _think_ an 576-byte MTU may be required for IPv6,
>since 576-byte MTUs are allowed for multicast traffic.
Last I looked they were going to require a 1500 byte MTU.
>and reassembly is the responsibility of host-systems in IPv6; hosts
>should use path-MTU discovery and fragment accordingly, but
>there's currently no good way to negotiate a path MTU over a dynamic
Reassembly has always been the responsibility of the receiving host in
IP. The whole point of path MTU discovery is to avoid fragmentation
in the first place. Besides, a sending host just doesn't send
fragments in any version of IP, unless it's IP implementation is
severly broken. Routers fragment IP packets.
Wasn't it the plan to rip out support for IP fragments from IPv6?
Christoph Badura firstname.lastname@example.org
You don't need to quote my .signature. Everyone has seen it by now.
Besides, it doesn't add anything to the current thread.