This one time, at band camp, Josip Rodin said: > On Sun, Sep 09, 2007 at 07:17:01PM +0100, Stephen Gran wrote: > > > > That's my point - that results in no database files on disk when clamd's > > postinst is run, resulting in it failing to start. > > Erm, surely people who purposely evade recommends can be expected to know > how to handle such a situation?
There are two things wrong with that statement: it has been until very recently the policy of apt not to install Recommends, and it is still a valid configuration option to set. Second, if the program will not start without these two files, then it is a Depends, not a Recommends. No amount of handwaving will change that fundamental relationship. I feel like we're talking past each other here, and we've lost a bit of the thread. The two databases, main.cvd and daily.cvd, need to be in /var/lib/clamav by the time clamd's postinst runs and the init script starts clamd. This makes it a Depends as the program will not run without them. The aptitude invocation you mentioned will install all Depends, and yet the program will not run. That is not an acceptable arrangement of the Dependencies, in my eyes. OK? Second, all of the proposed solutions (to something I am not inclined to think of as a real problem, but obviously YMMV) in this conversation have not provided a way that actually expresses this need without also installing and downloading duplicate data, and usually also result in a negative side effect worse than what we currently have. Consider all of your solutions at the outset: A user install clamav-data. Several times a day, they redownload the full data set, instead of only downloading deltas. A user installs freshclam: There are no data files at the time clamd is started. The daemon fails to work out of the box. A user installs clamav-data-bootstrap: They redownload the new data files at package upgrade time, just like now, and they also get no ongoing updates. Each of these is worse than what we have now, and each of these are what will happen after a single apt invocation in your proposed scenarios. A user will have to manually notice that this wasn't what they wanted, and manually intervene and install a second package later. Finally, if they then remove clamav-data in favor of freshclam, they actually remove the necessary databases in the process, making it worse again. > > That's not really acceptable to save a few megabytes of download on > > dist-upgrade. > > Well, that would be on every upgrade, not just dist-upgrade. With the amount > of security bugs that happen these days, that's a burden. I don't really agree that a 10MB download every few months is a major burden that needs backbends to work around, sorry. I agree that it would be nice to be able to not redownload 60MB of extra data a year, and if there was a clean way to do so, I think it should be done. So far, I don't see a clean way that makes it simple or better for the casual user. Unless you can come up with a way of making some package contain the database files, and be a hard Dependency such that apt always picks a working solution, but only for the first install on systems that will eventually have a working freshclam, I am inclined to think we're beating a dead horse. -- ----------------------------------------------------------------- | ,''`. Stephen Gran | | : :' : [EMAIL PROTECTED] | | `. `' Debian user, admin, and developer | | `- http://www.debian.org | -----------------------------------------------------------------
signature.asc
Description: Digital signature
