[Aboriginal] Website move and mercurial repository infrastructure.

[Rob Landley]

So the website is back on dreamhost, where it used to be in the
impactlinux.com days, and in fact impactlinux.com is now an alias
for landley.net so any old links you have lying around should work
again.  (Yay!)  And downloading prebuilt binaries should no longer
be amazingly slow.

I fixed the mercurial links to no longer have the name of the CGI binary
in them.  The apache rewrite rule fixes the web server to _accept_
links that don't have the name of the cgi binary in them, but it then
generates them relative to the CGI unless you do this in hgwebdir.cgi:

import os
os.environ["SCRIPT_NAME"]=os.environ["SCRIPT_NAME"][:os.environ["SCRIPT_NAME"].rfind('/')]

That tells it that the outgoing links it generates should be relative
to a base URL that doesn't include the CGI name, so the apache rewrite
rule can add the cgi name but the end users don't have to look at it.

This means right click and "copy" on generated URLs now produces
much easier to remember results:

  http://landley.net/hg/aboriginal/rss-log
  http://landley.net/hg/aboriginal/rss-tags

I note that the version of mercurial installed on dreamhost is an
older one that still has the "tarballs containing symlinks have the
wrong file length" bug, which doesn't affect extracting archives
but means that of you run "tar tvzf tip.tar.gz" it complains about
errors in the archive.  Sigh.  I can install a newer version of
mercurial in my home directory if it really bothers people.

My day job (doing kernel containers development for parallels) is a bit
time consuming, but on the bright side it gives me more incentive to
get the perl removal kernel patches upstream, and for getting the cron
job to regression test kernel builds up and running.

Rob