Re: Standards for WebServers
I propose to do the following:
- let the sysadmin choose his directory for ~www (yes, I prefer this
name over www-data; we have ~ftp, not ~ftp-data) even if we choose one
that we find would be the best;
- when a WWW package is installed, it can check for ~www and if it is
not the default ask if it should be changed (and warn that other
packages may need to be reconfigured);
This would allow people to setup for /home/www-data if they want,
or /var/web if their full /home is mounted, etc. The fact that
www is a `user' should not be a concern: *regular* (human) users
should be under /home, okay, but special users may be placed anywhere
IMHO (root, with /root, is itslef a good example).
We can then use whatever names you prefer for the document root, CGI
scripts and user dirs. I would like to stick with webspace, cgi-bin
and public_html because I like to have cgi-bin not be part of the web
hierarchy and public_html is history and standard.
Regarding the location of cgi-bin, I placed it in /usr/lib/httpd (same
for icons, btw) in my Apache package because /usr can be exported rw
to other machines with web servers sharing the scripts.
Mike proposed to have
/var/www/ cache
and
/var/log/www logs
If we want to use www, then I'd prefer
/var/spool/www/<package> cache for package
/var/log/www/<package> logs for package
I run Apache *and* CERN and don't want their logs to miss. In
addition, nothing guarantees that caches for two different packages
will have the same format, so it's not safe to specify a
package-independent location; as for the /spool/ component, all tools
I know who get/store/delete files do it in /spool/ (examples are
sendmail, lpr, uucp, news, etc.) because /spool/ indicate that the
files are under control of a program that will manage their
expiration. I'd like to keep this meaning for the proxy caches dirs.
Yves.
--
TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to
debian-devel-REQUEST@lists.debian.org . Trouble? e-mail to Bruce@Pixar.com
Reply to: