Bryan <(E-Mail Removed)> wrote:
> I want to determine when Windows XP / 2003 will use IGMP V3 vs IGMP
> V2 --
> supposedly the OS will detect the presence of an IGMP V3 network and
> use V3.
> In my own tests, this has been very problematic. So I need to exactly
> charactarize when XP will use which version of IGMP, including any
> associated timeouts.
>
Well that's a bit backwards: why use an old version if there's no-one on the
local link is using an older version. So the process is, use v3 unless
traffic of a previous version has been seen, if so then use that version.
Have you read RFC 3376? What differences have you seen between the
behaviour it specifies and the behaviour you have seen from XP?
> What we have seen is that V3 joins go out in the presence of network
> equipment that does not seem to understand IGMP v3. This is a real
> problem, and we need to understand it inside and out to support our
> customers, who are seeing failures related to this issue.
>
Well the typical kind of equipment that has problems with IGMP versions, is
IGMP-sniffing Ethernet switches. They work outside the IGMP specification
and it's thus not too surprising that they often mess things up.
They often are only capable of understanding v1, so if the routers and hosts
are using a later version they will be clueless! In all such cases, I'd
prefer them to stop trying to learn multicast destinations; they should just
flood all multicast traffic. I've not heard that they do anything like that
though. :-(
One of the threads I've taken part in noted an issue in the time between the
end-host booting and the router's first IGMP packet. In that time the
end-host, as per the spec, used the latest IGMP version. There's no other
behaviour defined in the RFC for that case.
--
Alan J. McFarlane
http://homepage.ntlworld.com/alanjmcf/
Please follow-up in the newsgroup for the benefit of all.
Similar Threads
Linear Mode
