[whatwg] Link rot is not dangerous
danbri at danbri.org
Mon May 18 04:45:57 PDT 2009
On 18/5/09 10:34, Henri Sivonen wrote:
> On May 15, 2009, at 19:20, Manu Sporny wrote:
>> There have been a number of people now that have gone to great lengths
>> to outline how awful link rot is for CURIEs and the semantic web in
>> general. This is a flawed conclusion, based on the assumption that there
>> must be a single vocabulary document in existence, for all time, at one
> The "flawed" conclusion flows out of "Follow Your Nose" advocacy, and is
> not flawed if one takes "Follow Your Nose" seriously.
> It seems to me that the positions that RDF applications should "Follow
> Their Nose" and that link rot is not dangerous (to RDF) are
> contradictory positions.
That's a strong claim. There is certainly a balance to be found between
taking advantage of de-referencable URIs and relying on their
de-referencability. De-referencing is a privilege not a right, after all.
If I lost control of xmlns.com tommorrow, and it became un-rescuably
owned by offshore spam-virus-malware pirates, that doesn't change
history. For nine years, the FOAF documentation has lived there, and we
can use URIs to ask other services about what they saw during that
Since there is useful information to know about FOAF properties and
terms from its schema and human-oriented docs, it would be a shame if
people ignored that. Since domain names can be lost, it would also be a
shame if directly de-referencing URIs to the schema was the only way
people could find that info. Fortunately, neither is the case.
> That link rot hasn't been a practical problem to the Semantic Web
> community suggests that applications don't really Follow Their Nose in
> practice. Can anyone point me to a deployed end user application that
> uses RDF internally and Follows Its Nose?
The search site, sindice.com does this:
"Yes Sindice dereferences URIs it finds in RDF instance data, including
class and property URIs. It performs OWL reasoning using the retrieved
information, mostly to infer additional triples based on subclass and
subproperty relationships. Doing this helps us to increase recall in
queries." (from Richard Cyganiak, who I asked offlist for confirmation)
Whether you consider sindice.com end-user facing or not, I don't know. I
put in roughly the same category as Google's Social Graph API. But it's
a non-trivial implementation that aggregates and integrates a lot of data.
BTW here's another use case for identifying properties and classes by
URI: we can decentralise the translation of their labels into other
languages. Here are some Korean descriptions of FOAF, for example:
More information about the whatwg