I completely agree with everything you just wrote, especially about Atom + APP being more than just a technology for blogs. APP is a great lightweight alternative to WebDAV, and promising for all sorts of data transfer. The fact that it has developer groundswell is a huge plus. During my Princeton days Kevin Clarke and I briefly talked about what a METS + APP metadata editing application could do. (I can't remember the answer, but I bet it would be snazzy.)
On the one side you are right: Atom + APP is becoming popular and the standards are good, so digital libraries should get into it. On the other side I was just reminded to the ECDL2006-paper "Repository Replication Using NNTP and SMTP": You can almost use any protocol (HTTP, OAI, ATOM APP, WebDAV, NNTP...) for most of digital libraries' use cases - but the best standard without approriate tools and support is pretty worthless.
I came to this realization out of frustration that most OAI toolkits (at the time, ca. 2005) didn't support that functionality well -- or at all. I don't know if that's still the case. However, the need to delete records is a reality for most projects, and OAI has somewhat awkwardly made us rethink how to "delete" a record in repositories and the like, both on the service and data provider end. You almost have to build your entire system around handling "deleted" records just for OAI exposure. In reality it seems like you just end up masquerading or re-representing its outward visibility on our local systems, which gets onerous. I guess the difference is that the growing number of Atom developers are heeding the requirement for deletions, whereas the few existing OAI toolkit developers have deemed that functionality as optional.
Most repositories do not even track deletions so they cannot syndicate them. If OAI-delete was mandatory, maybe OAI-PMH had not been used that much? OAI did a good job in promoting and documenting OAI-PMH but deletions were always treated as an orphan - I would not blame the standard but the lacking implementation. Also ATOM and RFC 5005 is not much better than other solutions - but its much more likely to get it implemented in Weblog and other software then OAI which is not that known outside the library world. Greetings, Jakob P.S: Maybe we would all be happy with Z39.50 if we had that wonderful Indexdata tools right from the beginning - instead there were only closed source specifications and different closed source partial implementations. A standard without easy to use open source implementations is condemned to be violated and die. -- Jakob Voß <[EMAIL PROTECTED]>, skype: nichtich Verbundzentrale des GBV (VZG) / Common Library Network Platz der Goettinger Sieben 1, 37073 Göttingen, Germany +49 (0)551 39-10242, http://www.gbv.de