https://bugzilla.novell.com/show_bug.cgi?id=288879#c4 Petr Baudis <pbaudis@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |pgajdos@novell.com --- Comment #4 from Petr Baudis <pbaudis@novell.com> 2007-08-21 09:13:30 MST --- Sorry for the delay, I took for granted that I've found the bug but I really didn't, and NSS isn't pretty to trace through. The problem is that currently, glibc takes ::1 /etc/hosts entry as 127.0.0.1 if you are looking up IPv4 instead of IPv6. And as apparently one of the few distributions, ::1 is associated with localhost instead of ip6-localhost on SUSE. Thus, nss_files "sees" 127.0.0.1 twice in /etc/hosts for localhost. What are our possibilities? (i) We revert the glibc change to take ::1 as 127.0.0.1. (ii) We rename IPv6 localhost record to ip6-localhost like most of other distributions, at the price of maybe a bit less smooth IPv6 integration (?) (based on some googling it just seems to create trouble for our users who don't care about IPv6 anyway, and IPv6-savvy users should find their way through). (iii) We get rid of the 127.0.0.1 entry and just keep ::1. "No-legacy" setup, but in theory might be confusing for the users. Things might get rough if you run app linked against old libc with libc + nss modules in custom LD_LIBRARY_PATH, and maybe in some other cases I'm not thinking of. (iv) We fix applications to expect duplicate addresses in getaddrinfo() output (I couldn't find anything saying that there must not be duplicate addresses there). I think obviously (i) is good only as a short-term measure at best. (iv) isn't overally very friendly. (iii) would be ideal but even if we decide to go for it we clearly can't do that for 10.3 at least anymore. So my preference is (ii). Your thoughts? I'm also Cc'ing /etc/hosts maintainer, and asked Ulrich why did he make this change at all. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.