Hilmar Lapp wrote:
It seems to me that domain-specific resolution systems are rather a fact
and we deal with them all the time.
We try to deal with it, but it's a pain, even though the number of
different systems I need to deal with is limited compared to someone who is
developing applications that must work across the entire life-sciences
domain, or even outside of this domain as well -- completely impractical!
For example, articles are referenced by DOI, entries in most
institutional repositories are referenced by Handles, and GenBank
sequences are referenced by a GI number. Any generic tool that wants to
deal with statements made about or to articles (presumably almost all
will want to) will need to know how to dereference a DOI. Alternatively,
for the time being we can prefix the DOI with http://dx.doi.org/ and
have a dereferancable HTTP URI.
That's the single best feature of that system, in my opinion :-)
I'm not sure why we can't apply the same principle to LSIDs. The life
science field isn't necessarily a small one, and it seems like a small
price to pay for a tool creator to implement a single resolution system
to resolve any life science identifier. Is this being naive?
From what I see, tool creators haven't shown much interest in implementing
domain specific schemes, or even at least make it easy to plug in your own.
How many semantic web tools support LSID resolution, for example?
There seems to be a notion that all "life science databases" will be
there in perpetuity, but in reality there are plenty of examples of
databases that lost funding and went "out of business", with PIR or BIND
being some of the better known ones. I'm not quite following why after
all these years of discussion the validity of URIs should again be
subject to the vagaries of funding, or the business acumen of commercial
enterprises.
The going out of business problem is a big challenge, but in my experience
the majority of changes are nothing else but URLs changing from something
like /cgi-bin/fetch.cgi?P00001 to /fetch.do?id=P00001 etc.
There are also some issues with such URLs that have nothing to do with
stability, such as the fact that there are no separate URLs for concepts
and their representations, see previous discussions on this list...
Domain names are quickly bought, used, and sold to someone else, and
this is not just theoretical. The proposed "ease" with which HTTP URIs
can be stably maintained first of all is clearly contradicted by the
empirical evidence that it's not happening right now (why would a W3C
recommendation change that? That we want stable HTTP URIs can't be new
to anyone), and second requires continued ownership of the domain name.
This seems like a trivial issue but in reality it's not once funding is
cut off.
For example, the journal Phyloinformatics discontinued recently and the
domain name phyloinformatics.org is now for sale. If they had used HTTP
URIs using their domain name, the next owner of the domain would
probably choose not to maintain any of those, or worse, reassign them to
something else.
What am I missing?
The time dimension? :-)
If you reference some resource on phyloinformatics.org, you do well to note
down the time when you accessed the resource (this is something most print
journals do when showing web addresses). This will later allow you to
retrieve the same page e.g. via the Internet Archive (if you are lucky).
Don't know how this is best handled in the context of the Semantic Web...