> But this isn't exactly what I'd call "mainstream" - there
 > isn't a hell of a lot of either IPv6 or multicast use today,
 > so I imagine that the multicast IPv6 use is an even
 > smaller set.

It's basically the same interface for IPv4 multicast applications, except
the kernel derives the interface index using the address that's provided
by IP_MULTICAST_IF.

 > That said, there should be a comment added to ip6(7p)
 > for IPV6_MULTICAST_IF, mentioning that the interface
 > index for an interface can change and if an application
 > is interested in this event, it should be on the lookout
 > with a routing socket. 

I'd much rather just admonish people against changing the interface index
in the first place, since there's really no reason to do it -- and indeed,
AFAIK, no one does.  So, again, I think all of this is a non-issue. 

 > See 6430864.

As per above, I think this is hopeless approach; there are simply too many
ways that applications interact with interface indexes -- requiring that
all applications that use interface indexes monitor a routing socket to
stay abreast of changes is untenable.

This whole thing seems like a "Doctor, it hurts when I do that" problem,
and I think the answer is simply "don't do that".

-- 
meem
_______________________________________________
networking-discuss mailing list
[email protected]

Reply via email to