- From: Eric Hellman <eric@openly.com>
- Date: Wed, 1 Oct 2003 12:05:10 -0400
- To: Patrick Stickler <patrick.stickler@nokia.com>
- Cc: <uri@w3.org>
Let me try to articulate one of the unspoken reasons in favor of the "info:" scheme with respect to the "http://some.root.domain/" approach. I think there is no argument that the http-some-domain approach has the great technical advantage over info that it has a built-in mechanism for retrieving resources and descriptions. What this list may not immediately recognize is that this technical advantage can, in many situations, be a great political disadvantage. In general, technological barriers are easily surmounted by technologist, but political barriers can be genuinely hard. By that I mean that getting people to agree on things (peacefully) is not easily arrived at by technological means. The technical ability to dereference an http URI results in a grant of power to the owner of some-domain by anyone who uses some-domain URI's. The practical effect is that real people, real organizations, and real communities are less willing to go along with the use of some-domain uri's to reference resources. Regardless of how much the owner of some-domain is trusted, there is a natural reluctance to delegate the power to de-reference. Potential users might try to impose specific dereferencing policies on the some-domain owner (for example by legal means) that might be objectionable to other users of the some-domain authority. The info scheme explicitly has no dereferencing capability. The lack of this ability makes it easier to get people and organizations to agree to use these uri's, and is in fact its signal advantage. At the same time, there is absolutely nothing to stop the proliferation of "info" dereferencing services. Given that hundreds of libraries have already installed OpenURL "dereferencing services" (such as our product, 1Cate (http://www.openly.com/1cate/) and others such as SFX, LinkFinderPlus, O-link, LinkSource, Sirsi Resolver, etc., etc.) which will rapidly add support for info uri's packaged in http urls, it seems likely that means for obtaining semantic web descriptions for info uri over http will be quickly supplied by market forces. To sum up, with regard to the info scheme, less is more. Eric At 3:06 PM +0300 10/1/03, Patrick Stickler wrote: >On 2003-10-01 14:45, "ext Jeremy Carroll" <jjc@hplb.hpl.hp.com> wrote: > >> While having some sympathy with Patrick here, I note that a new scheme that >> is URIs not URLs neatly sidesteps the car/document problems. > >Which is simply a problem that must be solved, if the Web >and SW are to progress... > >The tradeoff, having no easy means to obtain representations/descriptions >of resources denoted by info URIs is IMO far greater a loss. > >Particularly when the resources denoted are fundamental terms >used broadly in the classification of resources -- where the ability >to easily obtain schema based descriptions of those terms for >inference and other operations is a big win for applications. > >> An info URI identifies a concept not a document about that concept. >> A similar http URL would be taken as identifying either or both depending >> on who you talk to. > >True. But one of them would be wrong ;-) > >And that's why I've been working hard on URIQA, so that such >disagreements can be easily resolves by asking the web authority >for an authoritative description of the denoted resource, which >(hopefully) will include some rdf:type assertions. > >Rather than publish an insulated info: URI, an http: URI could >be used and precise descriptions of the denoted resources published >on the appropriate servers. E.g. > > MGET http://lccn.some.root.domain/2002022641 HTTP/1.1 > >where the server lccn.some.root.domain is managed by whomever >is responsible for managing the LCCN namespace, and a request >such as above would return RDF which explicitly defines the >resource denoted by that URI. > >Thus, the owners of the namespace themselves can *still* avoid the >document/car problem by employing a URIQA enlightened server >which provides precise descriptions of the resources denoted, >thus making it clear precisely what kind of resource it is. > >A new URI scheme that is not dereferencable is, IMO, a step >backwards (or at best sideways), not forwards. > >It's the proverbial "hiding your light under a bushel basket". >HTTP + URIQA offer a great table. Why not use it. > >Cheers, > >Patrick > > >> >> Jeremy >> >> Hammond, Tony (ELSLON) wrote: >> >>>> Why define and manage the URI space outside the scope of the core Web >>>> and SW machinery? >>>> >>> >>> >>> Hi Patrick: >>> >>> I have to query the question you put above. IMO the "info" URI scheme fits >>> full square within the core Web and SW machinery as articulated in the >>> latest Web Architecture Draft: >>> >>> Architecture of the World Wide Web >>> W3C Working Draft 27 June 2003 >>> >>> http://www.w3.org/TR/webarch/ >>> >>> The domain of URI is more extensive than HTTP alone. I would >>>assert that the >>> actual domain of URI is the Web. >> >> >> -- Eric Hellman, President Openly Informatics, Inc. eric@openly.com 2 Broad St., 2nd Floor tel 1-973-509-7800 fax 1-734-468-6216 Bloomfield, NJ 07003 http://www.openly.com/1cate/ 1 Click Access To Everything
Received on Wednesday, 1 October 2003 12:07:45 UTC