Le dimanche 10 décembre 2006 23:15, vous avez écrit : > Where did you get that syntax from? This is a fairly common getaddrinfo() hostname extension, present in at least GNU libc and BSD libc. > Why do I have to specify the interface if only one such link-local > address exists on the system? Because the IPv6 stack cannot know which interface to use for link-scoped address, since it could be about any interface (except lo and some fancy tunnels). > Or is it just a dumb linux kernel? It is made so *on-purpose*. That is not at all Linux-specific. > If it exists on more than one interface, why can't it then bind to > all interfaces that this address is valid on? I am not writing the IPv6 standards. I bet it's meant for consistency with the requirements for connect()/sendto(), and to avoid issues when using software bridges or similar constructs that may trigger multiple interface with identical MAC address and hence link-local IPv6 address. > And again: where is this documented? Partly in RFC 3542. Otherwise in glibc source code. -- Rémi Denis-Courmont
Attachment:
pgpoWEpYMoGk0.pgp
Description: PGP signature