On Sun, Jun 08, 2003 at 11:07:49PM -0400, Ned Ludd wrote: > > > I agree that having this would be worthwhile, however it has been said > > > higher up before that /home/httpd/htdocs would be where things were > > > installed. > What about the users that are changing there default DocumentRoot's? > Are we to ignore that completly and install to /home/httpd/htdocs > anyway? I believe that would be a better solution than looking for the DocumentRoot they have set. As an example, I run a couple of webservers at work. Their DocumentRoot is /var/www, which is a read-only NFS mount. I'd venture that those users than change the DocumentRoot are doing so for a very specific reason. One further thing with the DocumentRoot issue. Say I have two systems that share binary packages, but they have different DocumentRoot settings. I create the binary tbz2 on the first system, and it packages everything up with the first path, say /home/httpd/htdocs. Then I want to also install on the second system, which has a DocumentRoot of /var/www. The files would still be installed to /home/httpd/htdocs, defeating the purpose of your check for the DocumentRoot setting. What I do on my systems presently, (that don't have DocumentRoot as a RO nfs mount in /var/www) is let all ebuilds install to the /home/httpd/htdocs, and then symlink them into my own DocumentRoot location. On the same note with DocumentRoot, having Apache installed explictly implies a User of 'apache' and Group 'apache' existing. Rather than any magic on this, lets install all default packages with that uid and gid, and instead work on keeping that secure. If users change the User and Group setting in Apache again, they usually have a _very_ good reason for doing so, and so should be able to deal with the concequences themselves. > > > A few problems with this. > > > 1. What about people that have both installed? I actually have an open > > > bug for mod_php about this, and I'm trying to decide on a workable > > > solution. Presently I'm leaning towards an enviroment variable > > > 'FORCE_APACHE=1|2' that overrides any detection routines, for the > > > special case of people that have both. > Such and enviroment variable would make things quite less complex. Ok, could we possibly come up with something that everybody can agree on, and then use that everywhere. > > This is what the apache2 use variable is supposed to do. At least, as far as I > > know. > It does not behave this way for me, try USE=-apache2 with KEYWORDS=~x86 > while wanting to keep an apache2 from getting installed. The apache1/2 stuff on ~x86 is a mess, I strongly agree there. I'd really like it if apache1 just went away, but I see some users want it for various reasons. > Aparently the apache2 use flag was introduced to break the tie if both > apache 1 and apache 2 were installed but very few people are using it > correctly though, thus the need for these things to be addressed by an > eclass vrs everybodys the ebuilds themselfs. Having both installed simultaneously is a very messy business already, as they both use /home/httpd, and contain binaries with identical names. -- Robin Hugh Johnson E-Mail : robbat2@orbis-terrarum.net Home Page : http://www.orbis-terrarum.net/?l=people.robbat2 ICQ# : 30269588 or 41961639 GnuPG FP : 11AC BA4F 4778 E3F6 E4ED F38E B27B 944E 3488 4E85