Brian May wrote:
Its been a while since I looked at openca last, and I was looking at updating my Debian package.
However, I notice that OpenCA still has the braindead assumption that configure will be called once for every installation.
Under Debian this assumption is totally braindead, because the same package will be used on all systems. It is not possible for individual users to rerun configure, because that step has already been done by the package maintainer.
Using configure is OK to configure stuff like Unix UIDs, directories, etc. It is NOT OK to setup things like web host name, organisation name, country, hierarchy-level, etc.
A possible solution, the *.in files need to be distributed with the package (eg. in /usr/share/openca), and a script needs to be written that will only configuration *after* everything else is installed (eg. process the files in /usr/share/openca, process with sed, and produce the files under /etc/openca). Obviously configuration would only be for the first time, subsequent times would overwrite the initial configuration (please correct me if this isn't already a problem anyway).
Even better, if this script created a configuration file that could be used on all systems used by one organisation.
Previously I have done something for my Debian package, but it would be much better[1] if a solution could be maintained by you OpenCA people.
Comments?
Yes, we know this problem since a long time because it is a problem for our RPM-packages too. The problem is that we don't find a method which works on all systems - RPM-based, Debian, Solaris, AIX, FreeBSD, OpenBSD and NetBSD. We need a system like autoconf and configure which allow us to manipulate some files after the installation. A second installation is no problem. We can implement this easily with Makefiles.
The first idea was to create a second configure script. The installation would work like this:
./configure make make cutomize make install
But like everytime there are some problems. "make customize" is unusual but we can put it into "make install". The next problem is how we give the second script (make cutomize) the options? The next problem is that the user can't reconfigure the software after the installation if we install after make customize. Perhaps we should create *.in.in files and copy *.in during "make install" to OPENCADIR/config/ with some appropriate Makefiles to allow a reconfiguration after the installation.
So the usage could be:
./configure make make install cd OPENCADIR/config/ ./customize make install
The problem is that after the first "make install" the software don't work and the reconfiguration requires root permissions to change the owner and group during the second make install (but perhaps this is not a problem). So there a several problems and any comments are welcomed.
sed is perhaps a little bit problematical because every vendor uses it's own flavour. Another idea is to write a small Perlscript which reads a configuration from a file and then reinstall all changed files (in this case we have no problem with different seds). So we are waiting for somebody who has a great idea, knows a good software or has some new ideas for the above described solution.
Best regards
Michael
-- ------------------------------------------------------------------- Michael Bell Email (private): [EMAIL PROTECTED] Rechenzentrum - Datacenter Email: [EMAIL PROTECTED] Humboldt-University of Berlin Tel.: +49 (0)30-2093 2482 Unter den Linden 6 Fax: +49 (0)30-2093 2704 10099 Berlin Germany http://www.openca.org
------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ OpenCA-Devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/openca-devel
