- From: Mitra <mitra@path.net>
- Date: Tue, 28 Mar 1995 11:29:12 -0800
- To: pierre@indirect.com (Pierre Landau)
- Cc: uri@bunyip.com
At 3:15 PM 3/27/95, Pierre Landau wrote: >I've seen a lot of discussion recently on various URN schemes, but am still >puzzled as to some of the underlying assumptions of these models. > >Suppose that someuser@somemachine.someuniversity decides to "publish" a >document, be it an HTML one or otherwise. As things exist now, the user >tells friends about it, or puts a pointer on some HTML page to it. Ideally, >the user would register the document, obtain a URN, and distribute the URN; >other copies of the document would share the URN and therefore the access load. > >However, >1) Unless some encapsulation is used, how is a document to be tagged with >its URN? Suppose I copy a public-domain JPEG image of Jupiter from a server >which is very slow. I have its original URL, but not necessarily its URN, so >although by my having copied it I might make it available to others on the >net, they will not be able to identify it as the same object. There have been requests to have a URN tag as an optional header field in HTML. >2) URN search ability is vital. Suppose I'm writing a paper, and want to >refer to other papers. I need to be able to find the appropriate URNs for >those papers, even though I may know only author and title, but perhaps not >the journal name. All I've seen so far in the discussions is "some whois++ >server >will take care of it" which just begs the question of who will maintain the >database. Web crawlers such as WWWW and Lycos are only marginally useful in >this regard because the Web is expanding too fast and their search engines >are still not well enough developed. The "handles" system >(http://www.cnri.reston.va.us/home/cstr/handle-intro.html) addresses this by >allowing a flat namespace and distributing the load among several servers, >at the cost of requiring all updates to be centralized. This shouldnt be needed, if you can retrieve the document you are referring to, and if it has a URN embedded (e.g. in HTML header) then you have it. >3) Using the hierarchical DNS scheme for registering publishers works well >with departments at universities, but perhaps less well with a publisher like >joe@orion.tucson.az.us. Perhaps the Library of Congress or a similar >entity is a more appropriate "authority" to register publishers with. Under my scheme it works fine, orion.tucson.az.us registers joe.orion.tucson.az.us, I've no idea how it would work under any of the currently proposed schemes. - Mitra ======================================================================= Mitra mitra@path.net Internet Consulting (415)488-0944 <http://www.path.net/mitra> fax (415)488-0988
Received on Tuesday, 28 March 1995 14:32:04 UTC