Rob Sanderson wrote:
info URIs, In My Opinion, are ideally suited for long term identifiers
of non information resources. But http URIs are definitely better than
something which isn't a URI at all.
Through this discussion I am clarifying my thoughts on this too. I feel
that info URIs are especially suited for identifiers that are not only
long-term identifiers of non-web resources (an ISBN may identify an
'information' resource, but it's not a web resource), but also
especially when in addition all of the following are true:
0) Of potential wide-spread (not just local) interest. Ie, NOT a URI for
a record in my local catalog.
1) The identifier vocabularly itself pre-dates the web and was not
designed for the web. (ISBN, SuDoc).
2) There is not a controlling authority for the identifier vocabularly
that _recognizes_ it's responsibility to maintain persistence _and_ has
the resources to do fulfill that responsibility. That could be be because:
a) There is no single controlling authority at all, the control is
distributed, and they don't all have their coordinated act together for
a web-world.
b) The controlling authority hasn't yet realized that these
identifiers matter for a web world, and don't care about URIs.
c) There's nobody that wants to commit to this because they think
they can't afford it.
That's what I'm thinking. URI for a wikipedia concept from dbpedia?
Sure, use http. Those aren't going anywhere, because they are
web-native, they were created to be web-native, the folks that created
them realize what this means, and as long as their project exists
they're likely to maintain them, and they're project isn't likely to go
away.
URI for an ISBN or SuDocs? I don't think the GPO is going anywhere, but
the GPO isn't committing to supporting an http URI scheme, and whoever
is, who knows if they're going anywhere. That issue is certainly
mitigated by Ross using purl.org for these, instead of his own personal
http URI. But another issue that makes us want a controlling authority
is increasing the chances that everyone will use the _same_ URI. If GPO
were behind the purl.org/NET/sudoc URIs, those chances would be high.
Just Ross on his own, the chances go down, later someone else (OCLC,
GPO, some other guy like Ross) might accidentally create a 'competitor',
which would be unfortunate. Note this isn't as much of a problem for
"born web" resources -- nobody's going to accidentally create an
alternate URI for a dbpedia term, because anybody that knows about
dbpedia knows that it lives at dbpedia.
So those are my thoughts. Now everyone else can argue bitterly over them
for a while. :)
And yes, I agree fully that ALL identifiers ought to be expressed as
_some_ kind of URI. Once you've done that, you've avoided the most
important mistake, I think.
Jonathan
