The first attempt uses an XML file of photo meta-data, transformed with a simple XSLT on the client-side. This transformation is to a table of data with links to the raw jpegs, map positions (courtesy of Multimap) and years (thanks wikipedia). No relationships yet. Hardly a mashup, since the context data is linked only with URLs.
The second attempt uses Google maps. Now the map is embedded in the photo page, much closer coupling afforded by the use of Javascript. The step of including the Google Javascript source is coupled with authorisation via URL parameters- a neat application of the Shanley principle. The XSLT caused me a bit of pain here when generating the Javascript and it's still not right: the transformed HTML works fine in IE6 and if generated client-side with XALAN, but its no good on Firefox which just spins.
I now have to think how the students will use and develop this example: their server is inside our firewall and so Google won't be able to issue a key. They need a server in the DMZ for this I guess but we need that for the new Internet Applications Development module for September this year.