While working on a project for my day job, I cautioned against blind recitation of the publish-find-bind mantra and was asked for an alternative.� This followed our discussion of Danny's visibility write-up and so I scratched down some thoughts and exchanged a few emails with Danny for a sanity check.� Below is a capture of our exchange and the attached (apologies for the length) are the ideas as they stand after incorporating some (but not all) of the discussion.��
(Sorry for the Word attachment but (1) wasn't sure where to put it in the wiki and (2) it's too late to fool around with�uploading and positioning figures.� �Can remedy later.)
The email capture follows.� As typical, read from the bottom up.
I concur with your distinction between global and�groups.� It also does a nice job of mentioning�taxonomies.� You have captured the two forms of global�visibility that I would expect to see the architecture�discuss.� While an "aggregate of information expressed�in many vocabularies and categorized using many�taxonomies" is likely to be the more prevalent type of�visibility in a global SOA, significant efforts are�being made for the GIG architecture to "mediate�capabilities that enable discoveries across�vocabularies" as well as classification levels.
Danny,
Thanks for the quick feedback.� What I sent was�somewhat of a ramble�and a capture of thoughts without things completely�worked out,�without a clear idea of what points needed to be�made, without a�clear idea of how and/or where it eventually needs�to be�incorporated.� My following comments continue in�that tradition :-)
Global could certainly be looked at as a dynamic�group of groups.� I�am (and I'm sure most other people are too)�uncertain to what extent�there will be significant content generated�specifically for global,�so I'm not sure I'd characterize it exclusively as a�group of�groups.� However, that raises a possibly interesting�distinction.� As�a group of groups, global will be an aggregate of�information�expressed in many vocabularies and categorized using�many taxonomies�(or other representations).� Thus, either a global�registry search�will be akin to a search of the group(s) using that�vocabulary and�taxonomy (effectively, a search over a very small�part of the entire�registry content) or the global registry will have�to incorporate�significant mediation capabilities to enable�discovery across�vocabularies.� I think this could be a useful�distinction between�group and global and a driver for use of the global�registry.�Figuring out how it would be appropriate to express�this will be a�challenge and will require more thought.
Note, I think personal, group, and global somewhat�align with simple,�complex, and mediated, and those categories probably�influenced my�thinking but I was toying with concepts without
trying to�specifically match anything else.� No matter what�the categorization,�I think the boundaries may be fuzzy so, for example,�three people can�jointly maintain a personal catalog without fully�being a group and�three groups could work together without getting to�global.� As we�proceed I expect the boundaries to either become�more clear or get�permanently fuzzy.
The initial focus on UDDI and Web services had a lot�to do with the�actual work discussion that got me to start writing�this down.� I�agree that it needs to be expanded beyond UDDI but�that is the�prevalent straw man and somehow needs direct�attention because I�think the UDDI structure exacerbates the registry�problem.� Again,�this was in no way meant to be complete and your�comments are valid.�Again more thought is needed both on scope and�presentation.
Ken
On Jul 12, 2006, at 3:06 AM, Danny Thornton wrote:
Hi Ken,
The personal, group, and global categories make�sense�within the context of your document.� I can see�how�group can always imply some type of ownership�domain.�The question I ask myself is if global is another�type�of group.� If global is distinguished as a dynamic�ecosystem of groups, then that would make sense to�me.
�It would certainly be easy to integrate personal,�group, and global levels into the simple, complex,�and�mediated sections of Visibility. � I currently�have�one and two liner examples equivalent to the�personal,�group, and global examples. Your examples could be�substituted for those one and two liner examples.
The beginning of your document focuses on UDDI,�WSDL,�and Web Services.� I would also incorporate�discussions about ebXML, the Process Model, and�repositories.� See