On Thu, 2007-12-06 at 05:46 -0500, Dan Williams wrote: > On Thu, 2007-12-06 at 00:07 +0100, Tomas Janousek wrote: > > Hi, > > > > On Wed, Dec 05, 2007 at 09:12:28AM -0500, Colin Walters wrote: > > > 2) Know that you have to call res_init() in response to that signal > > > > > > Ideally we wouldn't have 2), but that's the state of things now. > > > > I do in no way want to start any distro flame or something, but just in case > > any of you didn't notice, I'd like to add that Debian has been in this > > "ideally" state for some time. > > > > They have a glibc patch that stats resolv.conf every time a resolving function > > is called. The resolv.conf file is managed by the resolvconf program and it's > > kept somewhere in /dev/shm. And, well, it simply works, all the time. > > I tried to suggest that a few years ago but upstream glibc didn't want > to do that, which I guess is fine. Maybe Fedora could just pick up that > patch, or start using nscd again by default. nscd wont help much for the simple reason the libc interface is very poor and forced many application to directly query DNS instead. For these application nscd server no purpose at all, while a caching server (dnsmasq someone suggest is the perfect fit) works with any app as it works at the protocol level. Simo. -- | Simo S Sorce | | Sr.Soft.Eng. | | Red Hat, Inc | | New York, NY | -- fedora-devel-list mailing list fedora-devel-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/fedora-devel-list