Chris,
you bring out my number one concern... xmlc was created (during Lutris'
consulting days) as an alternative strategy to "templating" approaches
like jsp and jolt/jddi. We learned a lot from building jolt (before
jsp was available) and using it in an early consulting gig at FedEx...
The target for xmlc was the interaction between html designers and java
developers. It's obvious that you and others are taking advantage of
xmlc's features to simplify that relationship. Another huge win was the
ability to support multiple presentation types (wml, html, xml, svg, and
even flash/xml). I'm concerned about a bottom-up approach to
re-thinking XMLC's internals might override what xmlc is all about and
what makes it a worthy alternative to jsp.
My recommendation to David, Richard and friends:
-> let's start a requirements document for xmlc 3.X and break it into
sections to deal with supportability/maintainability, developer, device
requirements and presentation designer requirements.
-> let's then make an announcement on theserverside.com for comments on
the requirements document, indicating that we're look for input from
those that deal with designer/developer interface in web application
development. I think it would serve as a nice way to resurrect a bit of
curiosity about xmlc. I know the folks at that site and they are dying
for content... it's pretty bad when they're featuring an interview with
their own people.
David
Chris Webb wrote:
Yep. It's a good feature because it allows our clients to create their
own look and feels and then test conformance against our reference
implementation. In most cases this means we do not have to be involved
in the process, which makes them happy and us even happier.
This process does need to be improved a little. At the moment the
compiler signals there is an incompatibility by stopping with the first
missing access method or return type and displays a message which is
fine for a Java developer but not to a mark up document designer. It
would be great if all incompatibilities were discovered in one pass and
presented in much more readable manner, i.e.
node 'foo' of type 'span' is missing.
node 'bar' of type 'link' should be of type 'span'
I don't know if it's possible to extract this information from the
compile process.
Chris
David Li wrote:
> Chris,
>
> So, you take advantage of the '-implement' option in the XMLC
> defining interfaces across different presentations, right? Hmm, I
> haven't used that for a long time but it's a good point on the
> strength of XMLC.
>
> I think this thread is getting interesting. All feedback and
> suggestion are valuable. Thanks and keep them coming.
>
> David
>
> On Thursday, Nov 21, 2002, at 19:04 Asia/Shanghai, Chris Webb wrote:
>
>> Mark,
>>
>> I agree with you on this. This change would be a completely different
>> beast to the current XMLC. If we lose the convenience methods then we
>> would lose the ability to create interfaces to abstract the dynamic
>> requirements of each document and thus compile time checking of
>> dynamic content compliancy. We use these dynamic content interfaces
>> extensively for multiple look and feels for HTML, WML and VoiceXML.
>>
>> Chris
>>
>> Mark Diekhans wrote:
>>
>>> Hi David,
>>> This sounds like a really interesting idea, however it's not really
>>> XMLC anymore, it's a different paradigm. It doesn't compiling
>>> anything.
>>> I would suggest that this would be a separate package. I think it
>>> could
>>> share the output and lazydom packages. This would also keep from
>>> break existing XMLC users, as this will be completely incompatible.
>>>
>>> It would be interesting to compare applications developed in this way
>>> with the current XMLC and Barracuda
>>>
>>> David Li writes:
>>>
>>>> DOM API is low level and was originally defined in IDL for cross
>>>> language portability. It's tedious to use and anyone who has tried
>>>> to populate a table using DOM API could testify this.
>>>>
>>>
>>> Yes, it didn't take long to figure out that a package like Barracuda
>>> was essentially for doing complex things.
>>>
>>>> The new XMLC reloading has open up the possibility to get rid of
>>>> the Java classes and enable to dynamically add new documents into
>>>> XMLC system without source generation and compilation. However,
>>>> with this approach, we won't have the convenient methods for the
>>>> XMLC class and will have to deal with generic DOM.
>>>>
>>>> There are two possible solutions to provide a new programming
>>>> interface on top of the XMLC: a Document centric approaches using
>>>> XPath or a Java centric approaches using either JDOM or DOM4J. Both
>>>> can probably be provided together and give the choice to the users
>>>> to pick their favorites.
>>>>
>>>> I have been experimenting with XPath for a while using a package
>>>> called JXPath from Jakarta project. It uses XPath to address the
>>>> elements in DOM and can be use to replace the convenient methods
>>>> generated by the currently XMLC implementation quite easily.
>>>>
>>>> for setting text,
>>>>
>>>> documentContext.setValue("id(foo)/text()", "New value");
>>>>
>>>> Just some random thought on the future of XMLC. I'd like to hear
>>>> what the community feel about the features needed to make XMLC a
>>>> better tools.
>>>>
>>> _______________________________________________
>>> XMLC mailing list
>>> XMLC@xxxxxxxxxxx
>>> http://www.enhydra.org/mailman/listinfo.cgi/xmlc
>>>
>>
>> --
>> Chris Webb
>> Voxsurf Ltd.
>> 3rd Floor
>> Elme House
>> 133 Long Acre
>> London WC2E 9DT
>>
>> Tel. +44 (0) 20 7240 3621 x 206
>> Mob. +44 (0) 77 8639 2359
>> Fax. +44 (0) 20 7379 7573
>> e-mail : chris.webb@xxxxxxxxxxx
>> Voice Demo. +44 (0) 870 744 7223 http://www.voxsurf.com
>>
>> Email disclaimer: This can be viewed at
>> http://www.voxsurf.com/disclaimer.html
>>
>>
>> _______________________________________________
>> XMLC mailing list
>> XMLC@xxxxxxxxxxx
>> http://www.enhydra.org/mailman/listinfo.cgi/xmlc
>
>
>
> _______________________________________________
> XMLC mailing list
> XMLC@xxxxxxxxxxx
> http://www.enhydra.org/mailman/listinfo.cgi/xmlc
--
David H. Young
http://www.kspar.net
1.831.246.1692 (cell)
| 