[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: about conformance
One of the difficulties we had in defining a conformance clause for
WS-Reliability, is the existence of diverse profiles of implementations: a
light personal device (e.g. cell phone) might only be required to act as a
Receiving RMP, and only be required to send acks. Or, a monitoring device would
only need to send, with the ability to resend. A message hub will need to act
as both Sender and Receiver, and support all features. It was hard to find a common basis of features to define conformance
levels on. But even so, it remains important to define conformance boundaries so
that implementations know where they stand on interoperability. An approach based on the QA guidelines (W3C / NIST,
http://www.w3.org/TR/qaframe-spec/ ) may help: That leads to distinguish: - Usage profiles, based on the different roles an implementation can
play: here Sender, Receiver, or both. - within each profile, functional levels (core, etc.) can be defined,
to which will correspond levels of conformance. For the sake of interoperability,
"core" Sender must be able to interoperate with "core"
Receiver, etc. - functional Modules can also be distinguished (a profile would require
the implementation of some modules, e.g. only {HTTP binding + resending
mechanism} for an HTTP Sender profile at Level 0, { HTTP binding + resending
mechanism + group management} for HTTP Sender level 1, etc. These definitions could belong in an (non-normative) adjunct to the
standard, something that helps developers characterize their implementations in
terms of profile/level (and also promotes a reasonably small number of
implementation profiles). That adjunct may or may not be merged later with the next release of
the standard. I propose we start discussing this in the meeting tomorrow if time
permits. Jacques |
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]