Hi Christopher,

first of all: sorry for the late response!! Many people were out all over
the last month (vacations)!!

> Howdy folks-
> 
> I've just today noticed the classes/interfaces
>   RequestedProperty
>   RequestedProperties
>   RequestedPropertyImpl
>   RequestedPropertiesImpl
>   PropertyName
>   PropertyParseException 
> 
> that have been added to our "main" package org.apache.slide.common.
> I hate to be a pain, but I have a couple of problems with the
> approach chosen. 
> 
> 1.
> The use of those classes is, AFAICT limited to the basicsearch 
> implementation. Apart from the search package, only the content 
> package has until now had a concept of properties 
> (== NodeProperty). So IMHO the above classes should go either 
> into search (preferred) or content.

You are right. We will move them either to the search or to the content
package.

> 2.
> The PropertyName, generally speaking, is a "(fully) qualified 
> name" (i.e. bundling namespace and local name). However, there 
> is already such a class in our repository: org.apache.util.QName.
> That class should be used and PropertyName removed, I think.

You are right. We didn't notice that class (...BTW, the class name "QName"
is a bit odd IMO...). We will use that class instead of PropertyName.

> 3.
> Both RequestedProperty and RequestedProperties are of such simple
> functionality, that splitting between interface and impl should
> not be necessary. I might be wrong here, but why should there ever
> exist a different implementation? Maintaining interfaces is 
> trickier than maintaining classes, so the use of interfaces 
> shouldn't be done without care.

You are right. We will get rid of the interfaces and have just classes ...

> 4.
> The implementations of those classes have clear dependancies on
> the WebDAV protocol (i.e. by parsing the JDOM element). This is 
> the probably most important problem I have with that code, and 
> it could be easily removed (just let the WebDAV layer do the 
> "parsing").

... and will do the parsing in extentions to that classes which will be
located in the WebDAV layer.

> I believe that, as a general tendency, the separation between 
> the kernel and the WebDAV layer has blurred quite a bit since 
> work on the the new features has begun (for example, how do I 
> access the versioning features without using the WebDAV code, 
> or knowledge about Delta V, for example from the taglib?). This 
> "blurring of the edges" has some serious impact on the overall 
> architecture, so I think we should pay close attention to it.
> 
> Thoughts, Comments?

Well, as we stated in our initial "Proposal for adding DeltaV support to
Slide", it was difficult to fully integrate DeltaV with the existing
versioning engine in Slide-kernel. We therefore decided for a
"semi-integrated" solution. As a compromise, we proposed to add a few
additional helper methods in Slide API (e.g. content package) for accessing
and navigating the DeltaV-specific history (...still to be done :-) ...).

Regards,
Peter

--
To unsubscribe, e-mail:   <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>

Reply via email to