|
ACS Home Related Links:
Advanced Campus Services |
Directory Services Project Purpose of meeting : Discuss status, review NSF NMI proposal, review tasks 1. NSF NMI proposal (see next page) (15 minutes) 2. Tasks related to prototype and Georgia Tech (15 minutes) 3. Other topics? ( 5 min) Next meeting : March 8, 2002, 3:30-5pm
Feb 19, 2002 email from Art to Vijay, Sham, & Mary Trauner: I would like to move forward on the NSF NMI proposal and have discussed with Mary Trauner some interesting approaches. It would involve resource discovery related to videoconferencing resources. Those resources might include: URLs, experts, directories, datasets. Mary is checking with the sponsor (UCAID) of her "commons portal" to see if there is a problem with using it as a component of the work (we can't imagine it would, but she is checking). The general concept might be: assemble around the portal the various resources; provide a mechanism like our clustering algorithm to "browse" those resources; and provide a mechanism for resource discovery. This last being an element that we haven't yet fully addressed, though it is assumed that at some point the clustering algorithms would run against some discovered objects. Resource Discovery is a current topic in the Internet2 Middleware VidMid working group (I am the chair...) and we've started laying out issues. Of course, UDDI < http://www.uddi.org/ > has come up, as has security (I may not want my objects "discovered" or only by certain persons...), and automation. I'd like to propose a title like: "Resource Discovery of video middleware objects using peer-to-peer for trust." Mary pointed out JXTA open source p2p < http://www.jxta.org/ >. At the recent I2 Campus Middleware Architecture Planning (CAMP) meeting in Tempe, AZ, Ken Klingenstein (I2 Middleware leader) was very excited about how p2p might be used. I'd like to consider how we could implement secure resource discovery by using p2p. It would have elements of distributed databases: distributing data the nodes of each user, permitting each user to set their own filters, even providing their own "access list" of trusted persons... Which takes us back into the "communities of interest" arena.
Feb 21, 2002 email from Art to Keith Hazelton, I2/U Wisconsin: As we discussed at CAMP, this describes our proposed approach to "promoting homogeneity" in the objects of directory space. It's interesting how the research focus helps bring some more practical things into the picture. As we build our prototype we ask: Where is the definitive specification of objects? There's a RFC for inetOrgPerson, eduPerson and comObject have source documents, but where are the similar specs for other objects? Do iPlanet, eDir, other LDAP implementations have the same way of describing objects? I kind of think that the object metadata varies slightly... or at least the level of completeness varies. Is there a need for a "resource discovery" provisioning service that provides a reference "mother lode" of object metadata? Would better "description" or "notes" or "semantics" information on creating an object promote better interoperability? i.e. can we show that better metadata improves object reuse and interoperability? (This works in human interactions: more discussion leads to more understanding ... if not agreement.) Is there a mechanism that permits already defined objects to be amended with additional information? Imagine there is a mechanism that permits "experts" to add annotations to existing objects, incrementally building up semantic richness. Especially wonderful if it could be done automatically - each use of object processed by a clustering technique that is constantly surveying the domain (Internet2 schools...?) and then re-displaying the "objects context map." ("Cool! Watch as that errant objectClass slowly moves to center as more and more schools implement it, or include it as the base for their extension. Neat! Let's us use it, it looks like a best practice winner!") In the Virtual Briefing yesterday on Multi-Campus middleware, the comment was made relative to objectClasses: As individual campuses create somewhat different extensions of objects for local needs, there would be some process for reviewing the various objects with the idea of bringing "good local ideas" into the "common set?" Exactly |
Last Updated: March 2, 2006