httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From (Robert S. Thau)
Subject Re: Content Negotiation
Date Sun, 05 Mar 1995 20:28:50 GMT
   From: (Robert S. Thau)
   Date: Sun, 5 Mar 95 20:01:32 EST

   I'm developing a sneaking admiration for the CERN scheme (scan the
   directory for files with the given name as a prefix; do type
   discrimination among those by suffix as usual) --- the overhead is a
   little more than I'd like, but I don't *think* it'll turn out to be
   prohibitive.  Besides, I still don't have any better ideas.

Somehow, that didn't come out right.  Must be the #$@!$ flu.

FWIW, it's easier to come up with schemes which have lower overhead
than the CERN scheme --- the gopher meta-file scheme, for instance ---
but I haven't yet seen one which didn't require fairly active
maintenance.  (The advantage of the CERN scheme is that you can
install a new form of a document simply by moving a file into place;
the server automatically adapts, so you don't have to recompile any
type-map files to make it notice the new version --- or worry about
version skew between the type maps and the actual file system).

One idea I was toying with was to have /some/multi-typed/page be
represented in the filesystem as a *directory* which would contain
*files* named html, html3, txt, or whatever.  However, this doesn't
save many cycles over the CERN scheme (the server has to scan smaller
directories, but a lot of the overhead of the scan is in opendir()
itself), and it seems to conflict with automagic directory indexing
--- the latter being a definite stumbling block.


View raw message