On Fri, Apr 15, 2005 at 08:18:10AM -0600, Patrick Walsh wrote:
> > - It doesn't work on the website mirror, which rsyncs the static html
> >   pages, but maybe that is a good thing.
> 	
> 	Definitely a good thing, I think.  
> 
> 	Why does it rsync when it could coda replicate them? :-)

When the mirroring was set up we didn't have realms and Coda wasn't
really stable enough to handle that kind of stuff. Nowadays it probably
will work, most of the data on www.coda is actually exported from /coda.
On the other hand I'm still getting some really funny disk usage numbers
when running 'df /coda' (1300% space used and things like that) so there
still are some problems lurking.

A mirror probably wouldn't need much more than pointing apache at the
static page tree, /coda/coda.cs.cmu.edu/project/www/html

> > - The server-side parsed pages are not cacheable by clients or proxies.
> 
> 	There are ways to cache dynamic content, but as a general rule it's
> understandable why this would be the case.  It probably isn't a big deal
> though since it will only be for the manual and not the entire site. 

Right now we're server-side parsing every .html page, I didn't want to
rename any pages to .shtml to avoid breaking links and such and the
XBitHack directive doesn't seem to be working. But it doesn't seem to
affect the CPU usage all that much, maybe we went from 1% to 2%.
Executing the comments cgi script on every page load is probably more
costly, but it shouldn't be hard to move it to fastcgi or something
similar if really needed.

The Coda manual pages should have the inline comments at the bottom.

In the short term we will drop docbook-sgml for the manpages and put the
nroff versions in the source tree. On a longer timescale we'd want to
update the rest of the documentation which is currently a mix of
linuxdoc and docbook-sgml, but I'm not really sure what the right format
would be. Docbook is just too verbose for hand editing, and the standard
seems to be a moving target. Is is still docbook-xml, or are we at -yml
or -zml by now?

Maybe we will simply use plain HTML text, or OpenOffice XML. As long as
editing is fairly simple and we can get reasonably looking .html and .ps
output.

Jan