cocoon-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Hunsberger, Peter" <>
Subject RE: Backend cache invalidation
Date Fri, 23 May 2003 14:15:44 GMT
Geoff Howard <> wrote:
> I've been waiting for this to come back up. I did a similar 
> implementation 
> in the past using JMS.  I made an avalon component that 
> subscribed to JMS 
> events that could then be published by some back end process 
> (in my case it 
> was java stored procedures called on update, etc in oracle).  
> The listener 
> then interpreted the messages and located keys in the cache.  
> Would that 
> work as a start?

I suspect so.  I'd love to have a look at it if you can.

> I could probably re-create that fairly quickly if you're interested.

We're interested!  

> Geoff
> At 09:42 AM 5/23/2003, you wrote:
> >It seems to me that this must be a classic Cocoon problem, 
> but I really 
> >haven't found any discussion of it or how to solve it:
> >
> >We have a generator that creates some content (from a database via 
> >EJBs) that for 99.99% of the time is unchanging.  Every once and a 
> >while an end user (with admin rights) will make a change to this 
> >content and immediately expect all users to be able to see 
> the changes 
> >upon a subsequent refresh of the relevant screen.  The content is 
> >expensive to generate and it makes sense to cache it.  We 
> could in fact 
> >cache it once and for ever if we didn't have the occasional update.
> >
> >Is there currently any way to have the back end signal to 
> Cocoon that a 
> >particular cache item needs to be invalidated?  We're thinking of 
> >building a MX Bean listener that would register itself with a common 
> >proxy super class for all EJBs.  Then upon successful insert 
> or update 
> >on the back end the proxy classes would signal the listener to blow 
> >away the particular cache entry.  Essentially this would be 
> something 
> >like a BackEndListenerCacheValidty (or BackEndCheckingCacheValidity 
> >although it doesn't really check) class.  Does this make sense? Does 
> >Cocoon or anyone have anything like this already?
> >
> >Peter Hunsberger

View raw message