[BYTEIO-WG] URGENT: answers to comments needed - Draft 0.4 of ByteIO Experiences document

[Mark Morgan]

> Which versions of the of the WS-Addressing Spec did you use?

http://www.w3.org/2005/08/addressing which corresponds to 1.0.

> Could you check Section 4b and comment on whether this is an adequate
> description of the process?

Seems fine to me.

> What is our agreed conclusion?

Well, I would say something along the lines of this:

The ByteIO interoperability fiesta success shows that the specification
describes a pair of port types which, with minor fixes as indicated in
this document, can be implemented by separate organizations in an
unambigous (with respect to interface and port type) way.  These grids
can, together with other specifications provided by the OGF (for
example, the OGSA-WSRF-BP), can then be used interoperably by users.
Further, the interop. fiest also shows that virtual interoperability
festivals are feasable and under the right circumstances can be used
effectively.

Its worth noting that the task taken on by Michel Drescher to provide an
interoperability test document which was then vetted by the standard OGF
document process proved invaluable in making the ByteIO Interop Fiesta a
success.  His careful attention to detail allowed for the fiesta
participants to provide rigorous tests that could easily be validated
and "graded" for success.

If there was any negative aspects to the interop fiesta, they would be
along the lines of the standards problems that seem inherent in OGF and
web services interop festivals in general.  Namely, the specifications
themselves rely on tooling, core specifications, and other 3rd party
products that tend to hamper success.  WS-Addressing and WSDL are both
complex specifications that have a tendency to promote only partially
correct implementations.  Fiesta participants often rely on tooling to
manipulate these specifications and end doing so can end up becoming
tied to a fundamentally flawed tool or library.  It has been this
writers experience that no grid implementation is free from this
particular problem and the fact that the issues seem to continually
raise their ugly heads indicates a fundamental flaw in either the
process, or the foundations on which the grid services world has built
it's specifications.

In summary, this particular interop fiesta has shown that ByteIO is a
reasonable and implementable specification that promised good
interoperability.  Many of the projects included have shown that the
specification itself is also useful to grid implementers and presumably
to their target users.

I believe we should cover:
- how doing the interop virtually was a good idea
- how having the interop doc from michel was a good idea, and how the
process of interop helped to improve the doc
- how tooling is a problem
- how WS-Addressing may throw up issues
- how we feel ByteIO will become useful in implementations

> Mark:
> 
> Could you contribute your part for Section 3c?

Done, please see attached.

> Could you elaborate on what aspects of WS-Addressing were a barrier to
> interoperation for Section 6d?

Done, please see attached.

> Can you check that the following: "The RandomByteIO interface has proven
> invaluable in the Genesis II system. Along with RNS, it is used through the
> system on every service implemented for management and user interface", is
> still true and accurate after my rewording.

Yes, I would say that that is a fair re-wording.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ByteIO_Experiences_Common_0_5.doc
Type: application/msword
Size: 281600 bytes
Desc: not available
Url : http://www.ogf.org/pipermail/byteio-wg/attachments/20080507/70cad8d8/attachment-0001.doc