- From: Al Gilman <asgilman@access.digex.net>
- Date: Fri, 24 Oct 1997 18:07:43 -0400 (EDT)
- To: leslie@bunyip.com
- Cc: uri@bunyip.com, connolly@w3.org
[At the moment, the trend would seem to be that URI stays.. so I am reducing the distribution a little because this is detailed stuff on the URL/URN harmonization agenda...] to follow up on what Larry Masinter said: > Some URLs don't accept relative references, e.g., 'cid:' and > 'mid:'. Maybe these should be URNs, too, but they're used > to locate a resource in a message, not to name it. So it's > fuzzy. > Various of the schemes now called URLs make more sense as names than as locators. And relative references such as 'ibid' and 'loc_cit' abound in the natural-language reference vocabulary we should try to serve in URNs. Referencing the previous article in a document published as a series of journal articles is a natural application for relative references in a namespace, as is referencing an article in the previous issue of the same periodical. > I think fragment identifiers that are used for *named* fragments > are useful in URNs and URLs equally. If fragments > were used as locators with some syntax "#bytes:1-47", we'd > have more of a problem. > [response 1 -- fragments as we know them] The general URL syntax draft wisely avoids trying to give a uniform concept or specfication for the the interpretation of #fragment substrings in URL references. Note that the "fragment" in this syntax symbol should therefore be read as "this fragment of the textual URL-reference," not "the cited fragment of the referenced resource" even though in the common case of http: URLs referring to HTML documents either interpretation fits. Given this level of caution in the URL spec, I see no way to claim that the concept doesn't translate into the URN domain. The concept is that you want to make a reference on a finer grid than the retrieval-chunking-factor of the resource. The object class of the object that you are naming can define what indexing methods it supports. You can use a #fragment phrase in UR* references for consistency and require that #fragment strings in schemes in the URN subseries are to be interpreted in accordance with published profiles for the class of the named object. The concept applies just fine to named information items. Whether you want to use the #fragment syntax is still up to you. [response 2, extensions pending] Both Dan Connolly and I have independently opined in the past that it should be possible for a third party to make a reference into a resource with a string-to-match, a line-number, or something objective that the resource author didn't necessarily mark as special. Line numbers are inappropriate in text/html but string matches are appropriate. Line numbers are fine e.g. in FORTRAN files where newlines are significant. A named object that is polymorphic across languages might not accept string matches but might accept section numbers. Or it might accept strings in conjunction with a language parameter. Etc. I still have hopes to see this kind of intra-resource reference refinement strengthened in the URI vocabulary of the 'Net. See for example my flame about "Where-it-says in URLs" at http://www.access.digex.net/%7Easgilman/web-access/wis_rfc.html -- Al Gilman
Received on Friday, 24 October 1997 18:07:53 UTC