Electric Forest

Electric Forest

thoughts about books, digital libraries, and stuff related to expressing and keeping track of our thoughts...

Wednesday, December 07, 2005

Is there a cure for my sinful Topic Maps thinking?

A while back I wrote an article about fixing really bad XSLT code, mostly digging into the functional and declarative nature of XSLT. One part if it talked about schema design, how this practice to me is a wonderful exercise in trying to get to the core of my data. I found that this sits in middle of the Data First vs. Structure First war.

So not that long ago, I wrote a personal email to one of the proprietories of the XOBIS specification, a library-grown XML format for better handling of bibliographical metadata. The mail was an asked for critique of the format seen through Topic Maps goggles, basically dipping into the whole "better semantics through fewer atomic bits" where you express models not through semantics of names of elements and attributes, but the actual contents of it, pretty much the core of Topic Maps. I explained how his entire schema can be represented in a simplified and elegant Topic Maps XML format, and Behold! there were converts.

With every new XML schema that comes along (and, oh boy, a lot of old ones as well!), it seems most people who create them fall into the pit of linking the semantics of their language into the data structures to represent their content. I've always felt this a bit back to front, and heck, one of the main reasons I turned to the Topic Maps side of things, even if I can recognize the temptation to do otherwise. So.

Back to the beginning. Whenever I come up with some schema of sorts to fix some internal or external issue, all problems looks like nails to me which I can bang my Topic Maps hammer onto. But we've been tought through time and credo that this is a very bad thing indeed, and that I should be ashamed of thinking this way. I should repent, but I can't help seeing any XML schema in a Topic Maps light and feel wonderment as to why they simply didn't just create a typified topic as a role-player to a given association-structure to solve ugly, clunky and complex constructions.

Unisex datamodels are daringly sexy, oh so sexy, but are there times when I should fight my lustful way of life and simply settle down with something uglier and be happy with that? I am awash with filthy thoughts.