Hello everybody,
I've got a complaint that a cache server under my administration
mishandles HTTP reget requests. I found nothing formally wrong about them,
but the way Squid handled the requests seemed to have some (odd) features.
Unfortunately, I had no time to read Squid's source, so I did a few
simple experiments using `client' and `tcpdump'. The results follow:
If Squid considers an object cacheable, but it encounters a cache miss,
then it fetches a whole object and returns a part of it.
If Squid is told not to cache an object using `no_cache' parameter
it uses `Range:' header in its communication with the server.
Everything is O.K. till this point.
However, if Squid is told not to cache an object using `Pragma: no-cache'
or similar HTTP directive it still tries to get a whole object.
The problem, to my mind, is that despites of imaginary RFC compliancy
of Squid's behaviour, a user has no way to do a real reget via Squid
in most cases.
I'd like to hear comments/corrections to the point of view on
Squid's behaviour. Thank you.
SY, Yar
--------------------------------------------------------------------
Yaroslav Tikhiy <yar@radio-msu.net> | Radio-MSU NOC | Moscow, Russia
Received on Fri Oct 09 1998 - 08:47:45 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:42:24 MST