Re: [Springframework-developer] Spring bean factory XML format

[Rod J. <rod...@in...>]

Juergen,

> I agree that the new format makes sense. I didn't think about validating
the XML before, but to my understanding you're right about the pitfalls.
Keeping backwards compatibility for the moment makes sense, as each and
every application context definition will be affected (admittedly in a
straightforward way).

Unless there are objections, I'll just check in the changes today. It's 100%
backward compatible, so all existing tests pass, and the existing test suite
covers all XML functionality.

I'll have to introduce support in properties format as well, but that's
really an enhancement, and doesn't affect existing use.

> So besides the <ref> tag, there's a <value> tag now, for mixed
collections. I guess it can also be used with a single value property like
this:
>
> <property name="name"><value>Rod</value></property>
>
> Do you recommend this syntax for such properties too? Does it add any
value in terms of validation? We should definitely stick to one recommended
syntax, to avoid confusion.

The new version accepts this, as well as the old form. I can't see great
validation superiority in the more verbose form. I agree it's best to have
only one approach. The only downside is the verbosity. I don't have any
strong views on this verbosity/consistency tradeoff. The <value> syntax
might be better in XML editors: you can see that you can have multiple
choices of ref or value elements within a property.

Also, Isabelle asked for a DTD. With the new format (overall) this should be
more meaningful.

> Regarding beans that expose CSV properties: I've already tried to clean
many of the exposed bean properties within Spring (e.g. both commandClass
and commandClassName, now only the former because of the ClassEditor), we
should try to continue this for multiple value properties. I guess if choice
doesn't add real value, it rather causes confusion.

Yes. CSV properties should go. I'll start with the AOP stuff.

> I'm not an XML expert, so I can't really help in terms of further
improvements. Does anyone else have some thoughts on this?

I think the "name" attribute may need to become "id" for validation, and my
new "ref" element maybe should be "href"? Anyway, it's trivial to change the
names of the XML, so I don't need to do this all at once.

Regards,
Rod