> I would like to propose that instead of (or if we must, as well as) allowing
> the embedding of information in the path (which is really poor!) we should
> special case the character ';' to mean end of URL for the client. This would
> allow folks to use
>
> http://machine/documentpath;info that can be passed to programs.
I think this is a much cleaner suggestion than the current scheme
>
>In the above case the server would return the document /documentpath, but keep
>the ';' information intact for scripts and programs. Forcing the server to have
> to stat each directory level is an incredibly waste of resources, and thats
> doubly so for AFS!
>
This was what I saw as problmatic also. But, in fact, that isn't the way
that NCSA httpd, at least, works (as Guy Descoux pointed out). What it
does is read some magic names (like "cgi-bin") from a configuration file
and assume that the first thing in the path after the magic name is a
script and anything else is info. It is not necessary to stat anything.
This is workable, but less clean than either Charles suggestion or my
earlier one (the maintainer has to set up and maintain the magic
names, for example). I also agree with Charles that embedding
information in a path is "really poor", primarily because it is
potentially confusing.
John Franks Dept of Math. Northwestern University
john@math.nwu.edu