- From: Martin Hamilton <martin@mrrl.lut.ac.uk>
- Date: Fri, 11 Aug 1995 10:10:28 +0100
- To: Ted Hardie <hardie@merlot.arc.nasa.gov>
- Cc: jg@w3.org, paulle@microsoft.com, http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com, janssen@parc.xerox.com, blampson@microsoft.com
Ted Hardie writes: | the Internet is over a very slow link. While the URN to URL resolution | proposals provide a useful model for long-term work, some very simple | solutions could work for mirrorinng non-interactive pages even in the | short term. A Mirrors: header, for example, could provide a simple list | of alternate URLS where the same data could be found (much like the lists | some ftp servers provide when user limits have been reached); the client Another approach which I've seen suggested would be to have multiple HREFs in an HTML anchor. This would be nice, in the sense that it would drop straight through the HTTP layer... Would it be kosher SGML, though ? Are there any plans to further develop the existing HTML & HTTP "HEAD" mechanisms for passing meta-information ? Methinks it would be good to have something along the lines of the "Dublin Core" element set [1] in there. Who knows, this approach might just make URCs redundant! :-) | could then analyze the list and choose mirrors which are closer. While | it may be difficult to determine what is "close" and what is "distant" | in network terms, there are some rules which could be built into the | clients. (Actually, a clever server could do the same thing with | redirects--read the client's network and domain from its headers and | issue a redirect to a closer mirror). The client could ping the servers and choose the one with the shortest round-trip-time ? Cheerio, Martin [1] <URL:http://www.oclc.org:5046/conferences/metadata/>
Received on Friday, 11 August 1995 12:13:18 UTC