Re: SEARCH by last path segment, Was: SEARCH for displayname

On Nov 18, 2003, at 4:36 PM, Lisa Dusseault wrote:

>
> Julian, you keep begging the question.  You effectively argue
> that putting properties on bindings is wrong because it's wrong.
> You can't assume that something is wrong in order to argue that
> it's wrong.  I'm asking you to consider whether our specifications
> would be better or worse if it *weren't* wrong.
>

I have to say that I'm with Julian on this... It's really a fundamental
principle of WebDAV that all properties of a resource belong to either 
the resource or to its parent collection(s).

If we can get the semantics of normal operation right using only one
abstraction (resource with properties), then I think that it's wrong to 
add another abstraction (binding with properties). Every example that 
I've seen in this discussion seems either more intuitive when done 
using properties in the parent collection, or seems like broken 
semantics that are going to cause significant pain later.

	-Mark


>>> If we defined a feature to hide bindings, you could set up binding
>>> 'foo'
>>> to resource A as hidden, whereas binding 'bar' to resource A as
>>> visible.  Then if you request 'ishidden' on 'foo'
>>> the server returns 'true', and 'ishidden' on 'bar' returns false.
>>
>> And this is exactly what I want to avoid. If the property
>> belongs to the
>> parent collection, nothing is lost and we don't need a hack
>> (a property
>> that varies upon request URI).
>
> Calling it a 'hack' assumes that we agree on the architectural
> principle.  I'm questioning the proposed architectural principle.
> Do you have another argument besides calling it a hack?
>
> Lisa
>
>

Mark Craig Chu-Carroll,  IBM T.J. Watson Research Center
*** The Stellation project: Advanced SCM Research
***      http://stellation.eclipse.org
*** Work: mcc@watson.ibm.com/Home: markcc@mac.com

Received on Tuesday, 18 November 2003 16:52:43 UTC