Mike Edwards wrote:
Luciano Resende wrote:
The targetNameSpace were only added on the Calculator and Chat
samples, as suggested on the JIRA 1330. The other changes were merely
cosmetic (such as alignments) or adding missing headers on some
composite files.
What's the suggestion here ? To remove the targetNamespace from the
Calculator samples ?
Umm, sheepish grin.....
This has revealed an erratum in the spec. The Assembly spec
contradicts itself by saying that the targetNamespace attribute on a
composite is optional (line 1033) but in the XSD and in the
pseudo-schema, it says that it is required.
Oh well, I'll raise this with the spec group and get a resolution....
My own opinion is that targetNamespace is essential only if there is a
desire to share the composite between contributions, since then there
must be a means to find the composite from a "remote" place. However,
other spec team members may prefer the "required" approach to avoid
errors....
I certainly was arguing to remove the targetNamespace from the
samples, since it had no use (didn't show anything useful in terms of
those samples).
On 7/19/07, Mike Edwards <[EMAIL PROTECTED]> wrote:
Folks,
I tend to agree with Ant. The namespace gets necessary once we get
into
inter-contribution sharing of artifacts and then the namespace is
essential, but I dont think it is necessary or useful in the simpler
examples.
I DO think we need to add the XSD location for the SCA namespace,
however - this is useful for tooling using the composite files, for
example. This is covered by JIRA 1443.
Yours, Mike.
ant elder wrote:
> Do we really need these, line 1033 of the assembly spec says the
> targetNamespace is optional, and the samples don't use the sample
> namespace? As they're samples isn't it good to keep the XML as
bare and
> simple as possible?
>
> ...ant
Yours, Mike.
I agree that the less we require in SCA assembly XML the better. However
I'm not sure that more options make the programming model simpler.
Actually, I had not realized that the namespace was optional, but making
it optional raises some questions:
- when do we recommend an app developer to use a namespace or not?
- how does an app developer reference a composite with no namespace in
<implementation.composite name="???"/>? I'm not talking cross
contributions here, I'm asking about the simple case where everything is
in one contribution.
- what steps does an app developer need to go through if he didn't
initially use a namespace and eventually changes his mind (because he
starts to use a composite)?
Are there simple answers to these questions?
With respect to repeating the location of the SCA XSD in all composites
to make some tools happy, I added a comment to TUSCANY-1389 [1]: The app
developer just needs to register the SCA XSD in his tool's XML catalog.
BTW I'm not seeing many WSDLs and XSDs out there referencing the
location of the XMLSchema for WSDL or XSD... but I've seen many people
do what I'm suggesting in TUSCANY-1389. I think it's better than asking
app developers to repeat the location of the XSDs all over the place in
their development artifacts to make some IDE tools happy.
[1] https://issues.apache.org/jira/browse/TUSCANY-1389
--
Jean-Sebastien
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
