Ross Singer writes:
 > I suppose my point is, there's a valid case for identifiers like
 > your doi, I think we can agree on that (well, we don't have to
 > agree, these identifiers will exist and continue to exist long
 > after we've grown tired of flashing out gang signs).  What I don't
 > understand is the reason to express that identifier as:
 > info:doi/10.1111/j.1475-4983.2007.00728.x
 > when
 > can serve exactly the same function *and* be actionable.

The problem with the latter identifier (and to be clear, yes, I agree
that it COULD function as an identifier) is that it gives the
impression that what you get when you dereference the DOI is that
specific resource, i.e. it enshrines as THE way of
dereferencing DOIs.

What if I don't want to get the article from  Maybe if I
go via that site, it'll point me to Elsevier's pay-for copy of an
article, whereas if I'd fed the DOI to my local library's resolver, it
would have sent me to Blackwell's version which the library has a
subscription for.  An actionable URI mandates (or at leasts strongly
suggests) a particular course of action: but I don't want you to tell
me what to _do_, I just what you to tell me what the Thing is.

Worse, consider how the actionable-identifier approach would translate
to other non-actionable identifiers like ISBNs.  If I offer the
non-actionable identifier
which identified Farlow and Brett-Surman's edited volume "The Complete
Dinosaur", it's obvious that you have a choice of methods for
resolving the ISBN; but if you offer the actionable identifier
then you are tying what should be a neutral identifier to one
particular fulfillment service.  How would Barnes and Noble, or indeed
your local lending library, feel about that?

Or, to get even more objectionable, what if I proposed the identifier
which not only points you to Amazon, but also causes money to be paid
to my affiliate account?  As soon as identifiers are dereferencable,
you've opened the system up to this kind of abuse.

(Now of course this doesn't settle the matter of what kind of URI to
use for an identifier.  If you decide you want something actionable,
then, yes, you pretty much have to use an http: URI, but if you want
something that is not, then you can use info:, or you can use http:
with a non-deferenenceable URL.)

 > I didn't grow bored with the argument, I just figured everybody
 > else had.


 _/|_    ___________________________________________________________________
/o ) \/  Mike Taylor    <>
)_v__/\  "Debugging?  Klingons do not debug.  Our software does not coddle
         the weak." -- Klingon Programming Mantra

Reply via email to