FYI, the rationale behind the use of identifiers for SNOMED-CT (ironically enough, given this thread) was to allow for some organizational semantics to be embedded in them. See: http://www.ihtsdo.org/fileadmin/user_upload/Docs_01/Publications/SNOMED_CT/SNOMED_CT_Identifiers_v1.0.pdf
[[[
The SctId does not contain semantic information related to the meaning of a concept or other component. It does however have a structure that is designed to allow different types of terminological components to be recognized. The nature of a component can also be derived from the table in which a component is distributed. However, the advantage of partitioning the SctId is that it also allows the nature of the identifier to be recognized when stored in a record or transferred in a message.
]]]
So, SNOMED-CT codes are actually not truly semantic-free identifiers. At least not in the sense of the OBO identifiers that we are discussing here.
Also, as a person who uses SNOMED-CT quite a bit, and builds tools that translate its distributions to and from OWL via various naming conventions, I actually find its use of codes to be quite a pain. Automatically generated Manchester OWL from SNOMED when the 'preferred names' are used instead make for a much more legible (and thus effective) way to quickly reference the meaning of SNOMED-CT concepts (even much more effectively than via the Clue browser, for instance).
Certainly, the risk that the preferred names change independent of the codes can be a problem for such a translation, but the use of the equivalent of annotation predicates such as skos:altLabel, rdfs:label, etc. can alleviate this - as I alluded to earlier. I guess this ability to have an effective annotation mechanism is an example of a difference between the SNOMED-CT representational framework and RDF (to answer your question about what makes them 'special').
--
Chime Ogbuji
Sent with Sparrow
On Tuesday, June 21, 2011 at 1:46 PM, Helena Deus wrote:
> Other standards (outside of semantic web) saw the need to rely on numeric identifiers, even if that created a burden for their users
> e.g. in SNOMED Lung =
>
> Of course it is a pain to query SNOMED with "all the diseases that affect T-28000".
> But the fact is that despite the inconvenience of having to fetch that identifier prior to the query, SNOMED is widely used.
>
> What is so special about semantic web identifiers that they don't need to follow the same path?