[RFC] hgweb: ease download of binary files

[Ted Pavlic]

> Already with Mercurial it was quite tricky, and I think I'll be
> stuck with 0.9.4 for quite some time on the server -- on the flip
> side, I've got branchview working, hehe.

Keep in mind that mercurial can be installed in your home directory on 
the server. E.g.:

make all
make install-home

Then, if you take a look at the hgweb and hgwebdir CGI, you can see that 
there are two commented-out lines that let you point to your OWN Python 
libraries. Uncomment those and change the paths, and you'll be up and 
running with Mercurial 1.0.

This process is what I used to get hg 1.0 up and running on my server. 
My server administrator was only comfortable installing the Mercurial 
package for his Linux distro, and that package was severely buggy (not 
hg's fault; totally the distro's fault). It was an easy fix to load 
everything at home.

Just a thought...

--Ted

P.S.

Regarding LaTeX, you still wouldn't have to source control all of that 
LaTeX build support (classes and styles, etc.). You just have to set 
your TEXINPUTS to look somewhere else.

In fact, you could use software (like Hudson) on your *OWN* machine to 
continuously build your PDF and then push it onto your server periodically.

OR, if you want to be a little more low-tech, you could setup a little 
script. Every time you're ready to merge heads in your central 
repository, build your PDF and put the date and changeset ID (e.g., 
69d51f99ddb2...) in the filename, and put all of those files in one 
directory. If they're sorted by date, you can easily go to past 
revisions. You could just display the directory on the webserver (turn 
Indexes on, like an FTP directory).


-- 
Ted Pavlic <ted at tedpavlic.com>