Jeni's Musings

I am relatively new (a couple of years) to the semantic web, rdf, etc, but do have over 20 years of experience in data (particularly, that which is relationally organized). The issue of URIs is not very different from that of how one would identify any data element (regardless of storage construct). As such, there is always an inclination to add business (i.e. meaningful) information into data identifiers. However, doing so almost always creates problems when changes occur (may of which have been described by Jeni in this blog post).

Given the problems I’ve encountered with identifiers that contain any meaning, my recommendation would be to utilize completely meaningless, randomly assigned identifiers (i.e.guids) for all URIs. In addtion to the problems meaningful identifiers create regarding change, most would argue that a minimal amount of information should be included in the URI to reduce the reduce the problems caused by the very issues discussed in this blog. Furhter, unless the URI contained a significant amount of information (which would be quite impractical), anyone wanting to gather information about the resource will need to see the properties that describe it; they are far more relevant than the URI itself. Quite often, many of the URIs are cryptic anyway.

So, to me, the best solution would be to utilize completely meaningless identifiers (URIs) for resources and let the properties/attributes assigned to them provide the meaningful information that is desired.