I've been making some time to think seriously about some of the assumptions and preconceptions I have regarding metadata, in light of my quick and dirty asset management tool.

I had committed 100% to the "rich, deep metadata is beautiful" mantra, drinking the IMS/IEEE LOM Kool-Aid™. I've built Large Applications that have been designed entirely around handling the complexity of this rich/deep metadata, and trying to abstract that enough to let me build actual functionality on top of it.

My epiphany over the weekend is that this approach is rather flawed. As King has been saying for almost 2 years now - the asset/learning object is the centre of things - not the metadata. And certainly not the rich/deep metadata specification.

I'm taking the opportunity to think critically about some of the projects I've done over the last few years, and am realizing just how simple many of them would be if I'd tossed the rich/deep metadata baggage long ago. For instance, the Teaching Resources collection was deployed as a test application, contrived to evaluate an XML database. It could have been built on top of a simple Dublin Core (or less) database, in maybe a couple dozen lines of PHP. I'll be rebuilding it this way (on top of my Dublin Core compliant asset manager) when I get a chance, and it will make it much easier to add new items, or to edit existing ones.

Thanks also to Scott Wilson, who just posted an entry "In Praise of Simpler Standards?" that nicely sums up some of the logic behind both ends of the metadata complexity spectrum.