Subject: Re: [tm-pubsubj] 1.A should URIs be name-based or meaningless?
* Patrick Durusau | | On the question: 1.A should URIs be name-based or meaningless? | | My gut reaction is that URIs, intended for human use, should be | name-based. I agree with that, but the question is: are all URIs really intended for human use? Some topic map applications are going to be creating PSIs for low-level individual topics and not just topics that are part of the ontology. I think that for ontology topics name-based URIs are right, because the name is part of the identity of the topic in most cases, and humans *will* be typing the names (the superclass-subclass PSIs from XTM are a good example of this). However, for topics like "school 22942" or "cultural event 929291" things are a little different. First of all, is any human being ever going to be typing the PSIs for these things? Secondly, what happens if "school 22942" changes its name? Should the PSI then also change, or is it OK for the PSI to be based on a name that's no longer in use? | Without belaboring the history of DNS, I think we all find it easier | to enter: http://www.ontopia.net as opposed its actual IP address. DNS also has another reason to exist: the IP address is dictated by the physical location of machines (kind of), so that if a machine ever moves it gets a new IP address, but the DNS name can stay the same. So DNS has the edge over IPs in stability, but I think for individual topics meaningless identifiers are more stable than name-based ones. I've had this discussion once already with John Cowan on the geolang mailing list, and his opinion was that the name-based/meaningless choice mattered less than having a stability policy. So I think probably considering the options (and the pros and cons of each), and deciding on a policy for when the PSIs change and when they don't may be the best route. | I think the same will be true for human authors who are using PSIs | in documents. Some applications with well constructed authoring | interfaces, may choose to use meaningless URIs for one or more | reasons, that are not displayed to the user. I think we can pretty much assume that applications will have authoring interfaces that hide the actual machinery. Of course, developers will still have to see the actual PSIs, at least for ontology topics, but I think creating recommendations primarily for people using text editors is not a good idea. (Not saying that's what you are advocating.) | Another point, which I am not sure Lars intended to raise by this | question, is that I would anticipate a practice of having the domain | name part of the PSI actually reflect that it is a PSI. I think we should ask ourselves why we want to do that. Does doing this achieve anything? If does achieve something, what do we do about those people who can't follow this practice because they don't have a domain (or can't create arbitrary machine names within it)? And, does this mean that *any* URI pointing to that host is going to be taken as a PSI, even if it might just be a navigational page or a stylesheet? (I did not intend to raise this question, but I think it's worth discussing. I think it should have a number of its own and be debated separately.) | For example, if the SBL were to start offering PSIs for biblical | topics, I would anticipate obtaining a domain name like: | psi.sbl-site.org. Does a couple of things: First, by custom | (hopefully in the near future) anyone using that domain name | recognizes that this is a PSI domain. Second, it clearly associates | the PSIs found there with a domain that hopefully the user | recognizes from other contexts as being the site from which they | trust PSIs. I think doing this is good practice, but I don't think we should require it, and maybe not even recommend it. I do think we should have some recommendations regarding the choice of URIs and so on, but we should be careful about putting meaning into the structure of URIs. | Continuing down the PSI string as it were, I would also anticipate | that the naming structure of the PSI would reflect something | recognized in the domain of its intended use. [...] You could do that, but I'd be very careful about taking that too far. It seems best to me to have a simple, consistent naming policy, and then to put the actual statements about the subjects into the topic map rather than to have them redundantly encoded in the PSIs. -- Lars Marius Garshol, Ontopian <URL: http://www.ontopia.net > GSM: +47 98 21 55 50 <URL: http://www.garshol.priv.no >