- From: Sampo Syreeni <decoy@iki.fi>
- Date: Sun, 10 Jun 2001 12:59:47 +0300 (EET DST)
- To: Pierre-Antoine CHAMPIN <champin@bat710.univ-lyon1.fr>
- cc: <www-rdf-interest@w3.org>
On 8 Jun 2001, Pierre-Antoine CHAMPIN wrote: >A problem is that the distinction is absolutely not taken for granted by W3C people. >Have a look at : > http://www.w3.org/DesignIssues/NameMyth.html If I understand correctly, TimBL's argument hinges on the problems we get from unstructured names and resolution. That is, he presents the idea that names change when they embed information (like people's names, metadata) that can change, and so the only sort of name that does not need to change is something which is completely unstructured. That is quite correct. Then he goes on to claim that without structure, names cannot be dereferenced. I think this particular conclusion is a bit hasty. There are a number of ways that enable one to deal with a large unstructured namespaces. One is just the usual hierarchical aggregation scheme that ISBNs, Ethernet address allocations and BGP tabulation of IP addresses utilize - we delegate blocks of addresses just as we delegate DNS domains. This time the delegation is not connected with the structure of the address, though, and can be changed at will. Although the above schemes mostly work to guarantee uniqueness, very little needs to be changed to enable resolution. It's just that the sort of model which results is different from what people are used to. Other examples of how to resolve difficult namespaces are hash based schemes, where the entire namespace is hashed and spread out real thin over multiple resolvers. Here one cannot necessarily know who will resolve one's identifier. So there are trust issues. Self-organizing search trees, as seen in some P2P block store architectures are a good idea, too: they even avoid single points of failure, like the DNS root. They might be fragile otherwise, though, and perhaps not efficient enough for widespread use. Still, it's easy to see that there are ways to resolve names which have all the properties one would expect of a URN. As I see it, there is no real reason short of convenience/short-sightedness to think that URLs should equal URNs. The distinctions between the two made by the URN working group are as relevant as ever, or perhaps even more so now that the Semantic Web is an issue. >Hence, I guess, the will of the URI WG to make the notion of URL obsolete. Probably. But once people start thinking about URLs as URNs, and abandon the effort that has gone to resolving generic URNs as opposed to the ease we can deal with URLs, I think the URN effort dies a horrible death. I think that URLs and URNs are two different things, neither of which is becoming obsolete very soon. On the contrary, people should work to get URN resolution working. Sampo Syreeni, aka decoy, mailto:decoy@iki.fi, gsm: +358-50-5756111 student/math+cs/helsinki university, http://www.iki.fi/~decoy/front
Received on Sunday, 10 June 2001 05:59:58 UTC