Re: Negative Caching wierdness with 1.1b10

cb@brewhq.swb.de writes:

>
>Hiya,
>
>When trying reload a page where some graphic icons had failed due to
>a network outage some 4 hours earlier I got the same "broken image"
>icons after a short delay and a TCP_EXPIRED_HIT/400 log entry.
>Opening the image file (via Netscape RMB) I was greeted with the
>"no route to host" error page generated by Squid, even though the

This bug is Squid's, and quite a serious one at that. We shouldn't
be sending IMS requests for failed objects....

>system was quite reachable. A reload produced a TCP_REFRESH/304 entry
>and still the same page, only a shift-reload caused a correct loading
>of the data. Any ideas what's going on here? (the negative caching
>values are both 2 mins on the server in question)

Blame for this one is less clear; maybe Netscape and maybe the HTTP
server. It has happened to me as well. When you press reload only,
netscape sends the no-cache pragma, *and* the "If-Modified-Since"
header. The caches pass on the full request (i.e. both no-cache and
IMS) and the server sends back '304 Not Modified'.

Duane W.
Received on Mon Oct 28 1996 - 10:51:39 MST