Forum OpenACS Q&A: Browser caching problem

Posted by Brian Fenton on
I'd be very grateful for any help the good people here could give me. I've got a client with 2 sites both running the same customised OpenACS. One installation has been running wonderfully for nearly a year and the client has been ecstatic in their praises. I only recently installed the second site and they have been having lots of weird problems, most of which appear to me to be caching type issues. As the software is identical on both sites, my deduction was that it was an infrastructural problem. This client's IT infrastructure is managed by a much larger organisation with (to me) an extremely complex setup and any conversations I have with the network manager end up leaving both of us frustrated at the other's lack of knowledge of our own area of expertise. ;-) I feel that I need to ask him
the right questions to help him give me the information that I need to solve his problem.

My question for you is: what are the links in the chain between the AOLserver and the user's browser where caching issues might arise? Firewalls, routers, VPNs, networks, PC OS, browser version, caching rules or settings on any of the above? Anything else?

The following might spark an idea for somebody reading this. This is what the client's network manager gave me as a description of the chain:
"Site2 Network - Netscreen50 Firewall - Cisco 2600 and into the hosting site as Site1 does?

All Clients access the system over the VPN using hosts files so no caching occurs here. This is a strange one as the only difference is the OS on each client, although we are now running 2 or 3 Windows XP machines in Site1 as well. The internet explorer is set to check for new versions of the page automatically, although like I mentioned earlier, all clients in Site2 have no DNS and no internet access, they only have what is allowed in the hosts file?"

He also added that all the PCs are running XP Pro SP1 with all patches and security fixes. And that "they are running some Group Policies to restrict some of what the user does but none of this messes with the standard browser settings."

Many thanks for all and any ideas.

Posted by Mark Aufflick on
Looking at, it is clear that the netscreen 50 is purely a firewall, so no caching is going on there. A Cisco 2600 won't be caching either.

I have seen IE on XP do some strange caching - but that's always been against dodgy C++ CGI's that I hacked together. OACS has always seemed fine.

One thing to try is to get one user with problems to try installing Netscape 6 (Mozilla, but that sounds less scary) and see if the problem persists. I don't know what to do of that fixes it, but more information never hurts. If you REALLY want more information - go onsite with a linux laptop and have fun with some packet sniffing tools!

You may also want to check this excellent thread on caching issues:

Posted by Brian Fenton on
Thanks for that Mark. That thread got me thinking about proxy servers. I guess in this case the VPN handles that... I must do a Google on VPN caching issues.

I have the no-cache on some of the more important pages already. I'll consider putting it in the default-master - wonder that'll affect performance? Shouldn't really I imagine as almost every page is dynamic already....

Posted by Bart Teeuwisse on

you could give the cache-control I wrote for OpenACS a try. See


Posted by Tilmann Singer on
If you have a 'logged in as ...' blurb on your pages, then you want them to have no-cache headers for sure.

Either put it in your default-master or backport the ad_http_cache_control from 5.0 and the occurances in the rp where it is called - this will result in pages generated without the default-master template to have the headers too. Or use Bart's proc - all a matter of taste ;)

Posted by Brian Fenton on
Thanks for the replies. I just want to clarify that this software has worked perfectly well on another site for nearly a year, so there is something up in the second site's infrastructure that's causing this. I should just put no-caching into the default master and that would probably solve the problem, but I'm also interested in learning something here.

One important fact I forgot to mention is that the caching is intermittent and not yet reproducible.

Thanks again.

Posted by Jade Rubick on