W3C home > Mailing lists > Public > public-xsd-databinding@w3.org > April 2006

RE: ISSUE-36: Tool selection for testing of basic pattern assertions

From: <paul.downey@bt.com>
Date: Sun, 23 Apr 2006 12:48:19 +0100
Message-ID: <2A7793353757DB4392DF4DFBBC95225504BFE906@I2KM11-UKBR.domain1.systemhost.net>
To: <public-xsd-databinding@w3.org>, <public-xsd-databinding@w3.org>, <jon.calladine@bt.com>

> Raised by: Jonathan Calladine
> On product: Basic
>
> On the call this week we briefly touched upon what our representative
> set of databinding tools would be that we tested against to formally
> prove the design patterns that would be included in the Basic Patterns
> document.

We've touched on this fairly contentious subject a few
times, including at our F2F.

I've shied away from a proposing a core set of tools not wanting to
risk saying '.NET 2.0 and JAXB are of course far more important than
anotherDatabindingTool just incase it turns out anotherDatabindingTool
is 'big in Japan' (by which I mean core to a large market place
outside of my personal sphere of knowledge).

Of course if we had a good representative set of vendors
participating in the working group, this task could
be a lot easier. Maybe your approach of listing tools important
to BT, or whoever, may flush out tools wanting to be on the list.

> This does not need to be a exhaustive list but for our own confidence
> it ought to cover many of the most popular tools across several 
> languages
> (its not just just .net and java out there....)

Again, 'popular' is the difficulty here, as is how far back
in time do we go - what if Axis 1.1 is critically important to
someone on the Working Group, or following our work?

The charter cites 'state of the art' but this term can be taken
several different ways - either the best available, the median
or lowest common denominator of features across all toolkits,
be they mainstream or niche.


> For discussion then, this is what I would choose currently to fit 
> our own
> core tools and those of our customers.
>
> Apache Axis 1.3 Final
> Apache Axis2 (latest 0.95?)
> BEA WLS 8.1 (clientgen and servicegen)
> BEA WLS 9.0 (clientgen and servicegen)
> IBM WAS 5.1
> IBM WAS 6.0
> JAXB 2.0
> JAXB 1.1
> XMLBeans 2.1

Can we pick the lowest version of these - based on a lazy assumption
that newer versions are likely to support more patterns?

> Microsoft .Net 2.0
> Microsoft .Net 1.1

and WCF?

> Axis CPP 1.6
> gSoap 2.7
> Rogue Wave Leif 2.5

OK ..

> SOAP:Lite

not a databinding tool (as PaulB notes).

> Oracle PL/SQL XML utilities (up for discussion but we have users of 
> this)

Oracle has a raft of toolkits, I'd be interested to see
an XML to Database mapping tool included.

This list seems to include toolkits used by BT, and possibly
tools used by well known customers, but may not really
capture the market place of tools potentially used by
customers. Hence the difficulty of this issue and process.

What about Ajith's WSO2 work or Peter's LMX? then there is 
Castor, JiBX, to name but three others.

> The rationale for the selection could well be the largest 
> commercial and open
> source tools for a language/platform. The multiple versions of 
> these tools in
> the list reflect the current user base as well as the latest 
> offerings in many
> cases. I think the list above needs to be reviewed and possibly 
> balanced with
> more scripting language tools.

What do we do about mainstream tools we don't have licenses
or environments to run - I don't have much personal experience
with SAP or TIBCO software, for example.

> The testing of our basic patterns with these tools justifies/
> validates the
> selection we make and in the border cases provides evidence as to 
> why certain
> xml schema constructs may not be present in the basic patterns doc.
>
> WI reliase that all of this may be contentious but what thoughts do 
> others
> have for a wish list of tools?

Yup, it's contentious, but maybe building a list of tools and
getting the community to provide more is the best way forward,
but what's our criteria for including or ignoring a toolkit,
especially one which lowers the lowest common denominator?

Is it useful to us or users of our specification if we elect to
ignore Axis 1.3 because it ignored xs:choice, even given how
widely deployed it is likely to be?

Paul
Received on Sunday, 23 April 2006 11:48:29 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Saturday, 18 December 2010 18:20:37 GMT