On Sun, 04.02.07 09:41, Bill Trost ([EMAIL PROTECTED]) wrote: > Interesting thought! When I first saw your email, I tried a few > experiments manually creating a static route to a link-local address > using a Linux version 2.6.18-3-amd64 kernel to see if I could make it > Do The Right Thing. Creating the route itself was pretty hard; "route > -6 fe80:..." kept giving an error like "unknown host", and adding that > address to /etc/hosts didn't help, either. Eventually, I was able to > use "ip -6 route add fe80:1::205:5dff:fef9:cc86 dev wlan0" to create > a route, at least according to "ip -6 route", but it still refused to > automatically use the right interface, even when wlan0 only had a single > address. I never did manage to create a route teling it to use XXX as > the source address when sending to the fe80:: address in question. > > Someone with a better touch for route commands might see if they can > make a miracle occur -- in particular, a route that uses a particular > address on an interface as the "gateway" might make sense, but I > couldn't get anything to accept such a route.
It's very unlikely that anyone is able to get this to work on current kernels. Have a look on net/ipv6/datagram.c:133 in the kernel sources. If no interface is specified for LINKLOCAL sockets, connect() will always return EINVAL. There is no workaround for this behaviour without patching the kernel. I am wondering if the kernel guys would accept patch to modify this behaviour to make link-local sockets more usable. Lennart -- Lennart Poettering; lennart [at] poettering [dot] net ICQ# 11060553; GPG 0x1A015CC4; http://0pointer.net/lennart/ _______________________________________________ avahi mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/avahi
