Explanation of purl.org/commons/ URIs
This is a work in progress.
Science Commons began exploring URI minting and resolution issues when
we realized that in order to generate orderly and durable RDF, we
would have to invent URIs for many things that don't currently have
adequate URIs, such as records in public bioinformatics databases.
The web architecture says to use
stable URIs, make them resolve, use different URIs for different things,
and use 303's for non-information-resources. We are
doing our best to implement structure this faithfully - it's not easy.
- Stable URIs: We use PURLs.
We don't have to establish the social of financial
infrastructure to maintain a dedicated domain.
Resolution of the URIs can change hands without the need to
- Resolution: PURLs for certain information resources can
forward directly to public sites such as NCBI. For everything else,
we redirect to a location that yields a 303, which in turn references
a script that generates a little explanatory page.
- Different URIs for different things: We're trying to set a
good example, and teach that a page describing a thing is not the
same as thing; the page and the thing need different names.
At least one other semantic web data integration effort suffers from
this confusion, and such a confusion is bound to lead to problems,
because computers are usually smart enough to handle puns.
- Use 303's: For anything that's not an information resource, and
for information resources for which resolution hasn't been
implemented in our scripts.
These URIs are still experimental; they are used in the
(including the sources used
in the IWWWC 2007 HCLS demonstration)
and are meant to stir up discussion. Please don't depend on their stability
right now, but do help with efforts that will lead to the creation of
stable and attractive URIs.
Databank records, considered abstractly, without commitment to a
particular representation. Particular representations include but
are not limited XML, ASN, and RDF.
HTML renditions of databank records.
XML renditions of databank records, using a predictable
format (e.g. DTD or schema); documentation pending.
PubMed records (not the articles themselves). Example:
Journal articles (not their metadata records), each identified
using the PubMed id of a PubMed
record that describes it. Example:
Articles may have other URIs as well. This one
can be identified by info:doi/10.1007/s00253-005-0186-4,
but unfortunately info: URIs are not unresolvable by most
Entrez Gene records (not the genes themselves).
MeSH records for descriptors and qualifiers.
(I'm not sure, but I suspect we may be mistakenly using
these URIs to identify the corresponding SKOS concepts, which would
This are misnamed, since they're not really records.
They are subject headings (SKOS 'concepts')
that are qualified versions of particular descriptors
(i.e., descriptor/qualifier pairs).
Science Commons URIs
Classes and properties belonging to the ad hoc
Science Commons ontology. Most of these are not information resources.
Classes and properties belonging to the MeSH ontology (from the
Why /commons/ vs. /science/?
The /commons/ URIs are meant to be administered under community
control, should that ever be seen to be necessary. (Currently they are
controlled by Science Commons.) Although they are not currently connected with
any independent entity, a promise to transfer control, sooner or later,
may be seen to be desirable in order to promote widespread adoption.
The /science/ URIs are meant to be vetted and administered by
Science Commons or an organization it designates. These should refer
to resources provided by Science Commons or defined idiosyncratically
by Science Commons.
We have tried to place URIs in the correct namespace as appropriate,
but have probably made a few mistakes.