Great that you are addressing this topic – it’s something I’ve been thinking a lot about myself the last couple of weeks, particularly the “time-varying property of something” scenario, like your population of London example. This is such a common pattern that I’m really surprised there isn’t already a ’standard’ way to do it.

So I look forward to the rest of your posts!

Bill

I have also seen the simple RDF and ontology examples that over-simplify models of the real world. Your London population example is perfect. One cannot, with any accuracy or precision, link together data from completely different sources without a common agreement of what those data mean.

Comparing recorded facts about “London” can vary based on:

- time
- definition of “population” (does it include all people who were there on a specific date? Who were residents? Extrapolated via an official census?, etc).
- definition of “London” (city proper? London area? Region?
- data usage (Has the data been adjusted for some specific reason? Has the data been adjusted for quality? Is there an accompanying legend or statistical reference? have they been rounded? By what method?)
- politics (are these numbers used in such a way that bias or other political gain might impact them?)
- source (did these numbers come from a census? a scientific study? a “welcome to London street sign?)

I believe that the lessons learned over the last few decades in traditional data modeling/data management can help deal with these questions. Ialso believe it is time for those showing how nifty it is to link up a bunch of data with none of the above questions answered to realize that they are demoing something quite different than they think they are.

Demos can and should be simple. But those doing the demo-ing should know what simplifications they have applied.