www-apache-bugdb mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Mark Nottingham" <m...@pobox.com>
Subject mod_autoindex/4520: mod_autoindex does not generate Last-Modified response headers
Date Sat, 05 Jun 1999 10:10:01 GMT
The following reply was made to PR mod_autoindex/4520; it has been noted by GNATS.

From: "Mark Nottingham" <mnot@pobox.com>
To: <apbugs@Apache.Org>
Cc:  Subject: mod_autoindex/4520: mod_autoindex does not generate Last-Modified response headers
Date: Sat, 5 Jun 1999 18:54:26 +1000

 > This is not practicable.  mod_autoindex only stats every
 > file for FancyIndexed displays, and it doesn't scan the
 > directory until after it has sent the response header.
 > Adding the behaviour you request would require that the
 > file scan be completed, and the Header and Readme files
 > be examined but not sent, before the response header is
 > sent back to the client.  In the case of very large
 > directories this could delay the response for a significant
 > interval, which is not good.
 
 I see your point. It's interesting; autoindex also doesn't send a
 Content-Length header, which rules out persistent connections (with both
 caches and browsers). I wonder where the break point between the cost of
 processing everything and the network benefits of full metainformation is?
 The gains made by persistance are fairly well-researched. What portion of a
 particular resource's client population comes through a cache is more
 slippery, but would be considerable for popular ones.
  
 A possible implementation might stat the directory to decide if it's small
 enough to justify the cost of computing the headers, or allow the user to
 configure whether they want them sent (possibly something like
 AutoIndexGenMeta or somesuch, defaulting to true, allowing people with very
 large directories to turn it off).
 
 Just something to think about (as if you guys don't have enough to do ;-).
 
 Thanks,
  
 

Mime
View raw message