On Wed, 2008-06-11 at 13:45 -0400, James Carlson wrote: > Peter Memishian writes: > > To that end, do we know if anyone actually depends on > > ifconfig working in this manner, or only the behavior of the ioctls that > > ifconfig uses working in this manner? If it's the latter, then we could > > tighten the ifconfig screws just to eliminate the usability issue. (Then > > again, there are so many usability issues with ifconfig -- hence the > > tentative plan for ipadm.) > > We already covered that ... PPP pretends to provide an Ethernet > interface to IP, so that (a) IP doesn't have to know anything other > than DL_ETHER and (b) IP doesn't have to learn about any system of SAP > values other than Ethertypes. > > Fixing just the user interface in ifconfig (and leaving the kernel bug > alone) would break the pppd "plumbed" option, while leaving the rest > working. (Nobody seems to know if doing that alone would hurt > anything else; X.25 was mentioned as a potential issue.)
OpenVPN's "tun" interface may also be a potential problem. I think it masquerades as DL_ETHER on Solaris (see http://openvpn.net/solaris/tun.c), but is setup as a POINTOPOINT IP interface when ifconfig configures the destination IP address. -Seb _______________________________________________ networking-discuss mailing list [email protected]
