On Thursday, February 8 2001, "Jonathan Rosenberg" wrote to "Eric Burger,
[EMAIL PROTECTED]" saying:
> Well, there is no such thing as a UDP MTU. There is an IP maximum packet
> size, which is 64k. The MTU is different on each link, and the path MTU is
> the minimum of the MTU along a message path. Should a packet arrive on a
> link which is in excess of the MTU, it is fragmented. THis is done at the IP
> layer, transparent to the applications. So, a proxy simply won't care that a
> 3k UDP packet arrived over ATM unfragmented, and will depart over ethernet
> with an MTU of 1500 bytes, requiring fragmentation. Fragmentation is not a
> great thing, as it has known error exponentiation effects, but functionally,
> its completely hidden from the proxy. Thus, the problem that has been
> described does not exist.
Except for all the embedded SIP devices which seem to have decided that they
don't need to support IP fragmentation. This has been a significant
interoperability problem at bakeoffs.
Needless to say, I think this is something that SIP implementors should
beaware of, and fix before considering their embedded devices
production-quality. I also think that we ought to add a torture test case
where crucial information in the SIP request sits after the 1500 byte mark.
Cc'd to sip-implementors; followups should go there.
--
Jonathan Lennox
[EMAIL PROTECTED]
_______________________________________________
Sip-implementors mailing list
[EMAIL PROTECTED]
http://lists.cs.columbia.edu/mailman/listinfo/sip-implementors
