Client side Toolkit and Reference Implementation

["Dennis E. Hamilton" <dennis.hamilton@acm.org>]

In thinking more about the architecture of reference implementations, it occured to me that there needs to reference client(s) as a check on the specification and certain questions around implementability of the specification.  I am not thinking of some sort of full-fledged client application, but a fixture that is used to confirm certain important situations:

1. That opaque elements that are used in turn-around operations can be used properly.

2. For elements that are not well-known -- left to implementations in the specification, understanding how a client can safely detect such elements and determine their significance in continued operation.

   2.1 Whether out-of-band agreements must be known in order to operate properly with such material, and how it can determine that the out-of-band agreements that are implemented into it are the ones that apply in a given situation.

   2.2 How, if there is material that is not well-known but is meant to be human understandable, how that works in having a round-trip operation be properly carried out.  (I am thinking of a situation where say, a policy name needs to be presented and the proper policy ID or other object returned based on user discretion or out-of-band knowledge wired into client-side software.

   2.3 In conjunction with (2.2), whether there is any part of the protocol for obtaining extended information that an user might wish to consult in making an informed choice concerning non-well-known material or protocol data elements.

3. What the assumptions are for deployment of user agents to the client side, how client side software is thought to be maintained, etc., and confirmation of those ideas using some test implementation.

It occured to me that a basic toolkit for client-side usage would be valuable not only for deriving clients that exercise and confirm how these conditions are to be handled.  It would also be a way to ramp up client-side tests of CMIS-enabled services and also be part of test fixtures that operate against CMIS-enabled services from the client side.

My thinking is obviously less than half-baked on this, but it is something that occured to me as I noticed the degree to which opaque protocol elements are passed back and forth and what may be required to cope with implementation-defined protocol elements.

 - Dennis

Dennis E. Hamilton
------------------
NuovoDoc: Design for Document System Interoperability 
mailto:Dennis.Hamilton@acm.org | gsm:+1-206.779.9430 
http://NuovoDoc.com http://ODMA.info/dev/ http://nfoWorks.org