[regrep] The case for an official Reference Implementation

[Farrukh Najmi <Farrukh.Najmi@Sun.COM>]

Team,

Here are some thoughts to seed the discussion on approving an
implementation as the official Reference Implementation (RI).

As some of you know I work for an organization whose main products are
Java API  Specifications.
The most relevant of these APIs to this TC is "Java API for XML
Registries" which provides a client API
for OASIS ebXML Registry as well as UDDI.

What Is Common in all these "Specifications"
---------------------------------------------

-Each specification must have exactly 1 RI that is used as a benchmark
for correctness and adherence to the specification.

-Each specification must have exactly 1 Technology Compatibility Kit
(TCK). The TCK has tests which certify compliance
and compatibility with the specification. An implementation must pass
the TCK before it can claim conformance to the specification.

-Each specification should have blueprints or best practices that
describe how to use it for solving common problems and needs

The RI is also what is used to develop the TCK tests with. The RI
provides a stable and reliable implementation for TCK developers so that
they don't have to develop the TCK against multiple environments.

The RI is also used as a data point and resource in case of spec
ambiguities to arbitrate disputes among implementations.

Characteristics of an RI
------------------------

-Not a commercial product

-Accessible to any one

-Design centered on strict adherence to specification, simplicity of
design and lack of external dependencies on other products

-Design not focused on industrial strength reliability, performance or
scalability (That is left to products)

-Commitment to keep up with future versions of the specifications

Why Should the TC Care About Having an Endorsed RI
---------------------------------------------------------

-Facilitates development of a TCK by providing a stable implementation

-Provides a venue for prototyping features of the next release resulting
in tighter specs.

-Valuable resource to other implementors common things like test code,
sample data and even code.

-Encourages more product implementations

-Encourages more client code development since it is available to
developers as a resource

-Ignites adoption

Why Should The SourceForge Implementation be the Chosen RI?
-------------------------------------------------------------------

If we decide to bless an implementation as our RI then
here are some reasons why this the implementation at:

    http://sourceforge.net/projects/ebxmlrr

should be considered as *THE* chosen RI:

-It is not a commercial product

-It is the only open source implementation for our specs

-It is free for anyone to use under an Apache like license

-Its design is centered on strict adherence to specification, simplicity
of design and lack of external dependencies on other products.
Its dependencies are all based on other open source projects.

-Design not focused on industrial strength reliability, performance or
scalability. Though it will be quite adequate for most users.

-This project's repository is already the keeper of the spec sources,
schema, SQL schema, UML
model etc., sample files etc.. Note that the spec related files are in a
separate module.

-This project is committed to work in close collaboration with a sister
TCK project or create a sub-project for the TCK.

-This project is committed to keeping up with future versions of the
specifications

-This project is committed to always being the first to implement the
latest specs.

-This project is committed to prototyping future features of next
version of spec
in our implementation

-This project is committed to helping *ALL* other implementations of our
specs with code,
advice and other help.

-This project's membership include almost 40% of the TC membership
(Suresh, Sanjay, Nikola, Farrukh, Joel Neu). Any one else who would like
to join should contact me.

-If anyone should have the status of officially recognized RI then this
project has earned its stripes.

-If not this project then *WHICH* project? The TC should consider this
project their project.

Summary
----------
In this email, the members of the TC are being asked to consider giving
their support to the need for an official RI and TCK for our specs.
Further they are urged to consider giving this honor to the SourceForge
project mentioned above. Finally I would like to ask the TC chair to
start a vote on this issue after necessary discussions and
deliberations.

--
Regards,
Farrukh