retrieveRevisionDescriptor and caching

["Whedbee, Marc" <ma...@gs.com>]

Hi,

We have noticed some behavior while using Slide 2.1 that we were hoping to
get some more information about.

We have been encountering problems opening files in MS InfoPath through an
HTML link.  Click the link once, modify and save the document, close
InfoPath.  Without refreshing the page, click the link again, the document
fails to open successfully.  An Ethereal trace showed that Slide was
responding to the GET request with seemingly-cached headers, resulting in an
incorrect content-length for the file.  The ETag generated for the content
of the "cached" response was also different from the kind we are creating
(we are generating the ETags as a timestamp so that they are always
different when the RevisionDescriptorStore.retrieveRevisionDescriptor()
method is called).  The response's Etag was
"f35c5048272409527de5a48535755490" as opposed to something like
"1111175691829"

Our logs also showed that the retrieveRevisionDescriptor() method was only
being called for the first click, not for any subsequent ones until the page
was refreshed.

We tried adding <parameter name="cache-mode">off</parameter> for the store
in our Domain.xml and  retrieveRevisionDescriptor() is now called for every
GET request, which solved the problem.  

Is there a reason why, with caching turned on, Slide does not still call the
RevisionDescriptorStore.retrieveRevisionDescriptor() method for every GET
request?  Or are we just  missing something?

Thanks.