Installation experiences…
I went to Koha's main site and found a download link. I downloaded and unzipped Koha 3.0. I found the newbie guide, http://www.kohadocs.org/newbieguide.html but it assumed you already had a working install. I didn't understand http://wiki.koha.org/doku.php?id=ubuntu_gutsy so I tried http://wiki.koha.org/doku.php?id=opensuse_11.0 That told me enough that I managed to add the IndexData repository, and got Apache and mySQL running and using UTF-8.
I had trouble with the CPAN Perl modules until I found that going into the Koha directory and using perl Makefile.PL told me what prerequisites I was missing, and then I could download them using perl -MCPAN -e 'install Missing::Prerequisite' They did need installing in a certain order, and http://wiki.koha.org/doku.php?id=ubuntu_gutsy then gave me the clue that I needed libgcrypt-dev and libgdm-dev. I had to install libgd 2.0.35 manually after much Googling of error messages, and tried several versions of CGI::Session before I found the right one.
After finding and fixing several more problems I decided I really must learn how to use perl and Apache properly someday. The bottom bit of http://wiki.koha.org/doku.php?id=ubuntu_gutsy gave me most of the remaining clues to get the system up and running.
I celebrated, set some system preferences, and then went to bed, shutting down the computer on my way. When I woke up and turned it on again, Koha wouldn't work. I eventually discovered that Zebra needs to run as a daemon, otherwise Ubuntu helpfully deletes the temporary files it needs.
Having got it running again, I tried to add a book to Koha. Apparently spaces in Z39.50 server names break things. Oh, and the British Library offers their data in a format Koha doesn't support. And the Library of Congress doesn't have most of my books, because I'm British and they're American. I eventually discovered the Oxford University server. I tried to import a MARC record. I thought I'd succeeded, but when I searched for the book in the catalogue, I couldn't find it. I discovered that if I typed in the book's barcode by hand, I could then check it out, despite the fact that it still isn't showing in search results.
The solution for this was to reindex zebra, using misc/migration-tools/rebuild-zebra.pl -b -w This showed up the problem that I never got export PERL5LIB working properly. The openSUSE page told me to put the lines into bash.bashrc.local, but actually they need to be in bash.bashrc on Ubuntu. However, the reindexing failed due to lack of permissions. Sudo rebuil_zebra.pl gave me the PERL5LIB error again. I found a suggestion of sudo perl -I/usr/share/koha/lib rebuild_zebra.pl -b -w on the internet, and that worked. Then I tried to add it as a cron job. After restarting Zebra, search now works! Thanks to Chris and everyone else for their help.
Now to try and add some more books…
You can find me on the Koha Users page on this wiki if you have any questions.