After installing the bind, bind-utils, and caching-nameserver rpms, the nameserver fails to cache, because /etc/resolv.conf does not have a 127.0.0.1 entry. I think it would be helpful to include a post-install script in the caching-nameserver rpm, which would (1) cp -p /etc/resolv.conf /etc/resolv.conf.old (2) add "nameserver 127.0.0.1" before any other nameserver lines in /etc/resolv.conf Thanks for your support.
Ok, I see this was reported previously as bug 2806, to whihc I found the following reply: // begin included text ------- Additional Comments From jbj 05/15/99 16:23 ------- The installer (and ultimately the user) is responsible for configuring the DNS client. Assuming that there is a nameserver running on localhost is not the Right Thing To Do. // end included text I respectfully agree AND disagree. Assuming that a DNS is running on local host is the Wrong Thing. But ignoring the possibility is also wrong, and not not particularly helpful. The Right Thing To Do is handle it. It is emminently reasonable to expect that someone installing "caching-nameserver" has also installed "bind." But, you say, what if they didn't? Easy. The post-install scripts of *both* "caching-nameserver" and "bind" check for the pre-existence of the other. If either finds the other and does not find "127.0.0.1" in /etc/resolve.conf, then it edits resolv.conf (after making a copy). There might be a pathological case I havn't considered, in which case the post-install script(s) should at least write a diagnostic to /var/log/install (or similar) to alert the semi-savvy admin that he needs to do something. \begin{personal_opinion} I'm reasonably experienced at this (pre-RedHat days, but still not a self-proclaimed expert) and it took me more time than needed to figure out what was wrong when my connects were slow. I consider this a user-acceptance issue. I'm persistent. But as Linux reaches for the mainstream, we all need to be aware that some people have a low threshold for frustration that *seems* unnecessary. Networking issues may be a bit much for a newbie to bite off, so we should help them whenever possible. \end{personal_opinion} Again, thanks for your support.
While I agree with all that you are saying, I do not believe that rpm scriptlets can be written robustly enough to handle the myriad problems that have to be handled to do the Right Thing. For starters, consider that the insaller environment in which the scriptlets run is chroot'ed w/o any running daemons. So, for now, let's just say that we're not gonna be able to address the problem.