[ILUG] Squid vs Browser cache

[Fergal Daly]

At 08:38 19/02/00 -0500, Subba Rao wrote:
>The squid access.log show the following:
>
>950966673.946    501 127.0.0.1 TCP_MISS/301 530 GET 
>http://pws.prserv.net/truemax root DIRECT/pws.pr
>serv.net text/html
>950966675.095   1098 127.0.0.1 TCP_MISS/200 4680 GET 
>http://32.97.166.75/truemax/ root DIRECT/32.97.
>166.75 text/html
>
>The squid store.log show the following:
>
>950966673.946 RELEASE FFFFFFFF  301 950966342        -1        -1 
>text/html -1/300 GET http://pws.pr
>serv.net/truemax
>950966675.095 RELEASE FFFFFFFF  200 950966239        -1        -1 
>text/html -1/4505 GET http://32.97
>.166.75/truemax/
>
>The above pages are static. But it appears that Squid is downloading it 
>again on revisits.

I'm not 100% sure of what follows, but at least you can test my theory. 
They may seem static, but the web server is not saying that have a look at 
what is returned when the page is requested

fergal at host:~> telnet pws.prserv.net 80
Trying 32.97.166.75...
Connected to pws.prserv.net.
Escape character is '^]'.
GET /truemax/ HTTP/1.0

HTTP/1.1 200 OK
Date: Sat, 19 Feb 2000 13:54:37 GMT
Server: Apache/1.3.9 (Unix)
Connection: close
Content-Type: text/html

fergal at host:~> telnet www 80
Trying 194.145.128.11...
Connected to oompa.esatclear.ie.
Escape character is '^]'.
GET / HTTP/1.0

HTTP/1.1 200 OK
Date: Sat, 19 Feb 2000 14:04:07 GMT
Server: Apache/1.3.11 (Unix)
Last-Modified: Mon, 27 Sep 1999 15:54:36 GMT
ETag: "401b-3a5-37ef933c"
Accept-Ranges: bytes
Content-Length: 933
Connection: close
Content-Type: text/html

the difference is that the esatclear one returns a last modified header.So 
if squid is asked for the page, it can check the change time instead of 
downloading the whole page. The truemax page gives no mtime and no caching 
instructions so presumably squid feels obliged to get it everytime. To see 
if my theory is correct, point your browser at www.esatclear.ie a couple of 
times and see what happens. Let me know how you get on.


> > I'm not sure what DB you're taling about, do you mean the cache of stored
> > web pages? I haven't ever seen squid set to do updates at set intervals.
> > Whenever I have set up squid, it followed the standard HTTP caching rules
> > ie. look at the headers (and possibly meta tags, not sure) for details of
> > how long this to keep the page for before it must reload it.
> >
>
>I was talking from my observations in cache.log. It appears the stored pages
>are done on an hourly basis (apporximately).
>
>2000/02/19 02:15:54| NETDB state saved; 134 entries, 4 msec
>2000/02/19 03:03:07| NETDB state saved; 134 entries, 344 msec
>2000/02/19 04:06:19| NETDB state saved; 134 entries, 754 msec
>2000/02/19 05:00:18| NETDB state saved; 134 entries, 12 msec
>2000/02/19 06:00:29| NETDB state saved; 134 entries, 314 msec
>2000/02/19 06:49:11| NETDB state saved; 134 entries, 404 msec
>
>If the pages are not stored then, would squid go out again and get it?
>That would be a waste of bandwidth.

I think cache.log is actually squid's internal state database, it's what 
allows squid to remember what pages it had saved etc. the last time it was 
running. It does not actually contain stored pages, these are contained in 
files in the cache directories. Cache.log tells quid which file corresponds 
to which web page. Calling it .log was a bit silly when it's really a 
database index.

Fergal