Re: [squid-users] Squid 3.0

> On Tue, 4 Jan 2005, Konrad wrote:
>
>> 1) It seems that squid HEAD version (squid-3.0-PRE3-20050103) doesn't
>> compile, out of box :), with esi-enabled.
>> There is a conflict between expat(1.95.7) and libxml2(2.6.16) parser -
>
>
> We know. How the parsers plug into Squid needs to be modularized to
> allow the parsers to coexists happily.
>
>> redefinition of XMLCALL macro. I've attached small, rather tricky,
>> patch fixing this problem. There is also one strange thing with
>> libxml2. It installs header files in PREFIX/include/libxml2/libxml,
>> and upon making symbolic link libxml->libxml2/libxml, in the include
>> directory, compilation fails. Is it bug in libxml2 ?
>
>
> Looks like a bug in libxml2 to me.
>
>> As I understand, this is the only caching rule which should apply on
>> ESI entities, but I've noticed that squid reacts also on ETag, Date,
>> Last-Modified, and Age headers from server, and Cache-control from
>> client.
>
>
> Probably correct.
>
>> Scenario 1:
>> mozilla -> squid -> Jboss with Tomcat
>> Forcing squid to cache ESI entities requires setting "Date" header
>> with value set to max-age seconds in future, so the respons from
>> appserver needs to be something similar to:
>> HTTP...
>> Surrogate-Control: max-age=SOME_TIME
>> Date: THIS_MOMENT + SOME_TIME
>> ...
>
>
> You should not need to mess with the Date header. In fact it SHOULD be
> set to "Now".
>
>> Tomcat by default adds ETag and Last-modified headers when it serves
>> static content, ie image fileas, *.css, or *.js. To convince squid to
>> cache those files I need to "zero" ETag and set Last-Modified into
>> future - similar as with Date.
>
>
> Same here.. You should not need to mess with Last-Modified.
>
>> Scenario 3:
>> mozilla (Ctrl+F5) -> squid - enforcing page reload with fresh content
>> Squid retrievs new content even though it have not yet expired.
>
>
> Probably true.
>
>> The question is, are these behaviours correct ?
>
>
> The date problems certainly is not correct.
>
> Regards
> Henrik
>
Thanks for replay

Regards
Konrad
Received on Wed Jan 05 2005 - 02:10:02 MST