Tim Bray writes all kinds of reasons why atom:id is not good... Ok. So, let's rename it to: "atom:uri"... Think of the atom:id/uri as the URI for the entry. Given that an entry is a distinct resource, all sorts of restafarian philosophy should support this. Expressing it in this way would also encourage people to follow the "proper management" methods that are necessary to ensure that atom:uri's are unique. Just as people do what is necessary, (they have little choice), to ensure that the "URL's" for their webpages are unique, they'll understand that atom:uri's should be unique. With atom:uri, we could start to provide services similar to what we do with news. News dealt with the "duplicate" item problem some time ago due to the need to identify items that were cross-posted to multiple news groups. People didn't like being forced to read the same posting in multiple groups and wanted to be able to tell their readers that a cross-posted item should be marked "read" if any of its instances had been read. Support for Message-ID was also introduced in TBL/Masinter's old URL RFC. For instance, even if an item has been posted to 10 different newsgroups (or feeds...) I can refer to it in a group/feed-independent way and retrieve it using an RFC1738[1] defined URL such as "news:eA#sg11#DHA.2664@xxxxxxxxxxxxxxxxxxxx"; which uses the news "Message-ID" much like a URI would be used. If atom:uri is supported, I could retrieve a specific atom entry using something like:[2] atom:urn:newsml:iptc.org:20001006:NewsMLv1.0:1 where "urn:urn:newsml:iptc.org:20001006:NewsMLv1.0:1" was the URI (an URN)... This would allow people to link directly to entries, without having to use an intermediary web site or alternate resource. It would also make it much easier for feed processors to recognize multiple instances of the entry. The problem of duplicate detection and group/feed independent identification of entries was long ago recognized, addressed and resolved in the NNTP space. The problem there, posts to multiple newsgroups is very similar to the problem of entries in multiple feeds. Certainly, the NNTP Message-ID system doesn't work perfectly, however, the large number of news readers and other utilities that properly create and handle news Message-ID's indicates that it should be workable and useful to support something similar in Atom. bob wyman [1] http://www.ietf.org/rfc/rfc1738.txt [2] The "NewsML" URN is defined by: http://www.ietf.org/rfc/rfc3085.txt
Attachment:
smime.p7s
Description: S/MIME cryptographic signature