Re: Compound Locations

On Wed, 21 Sep 2005 03:10:57 +0200, Nick Kew <nick@webthing.com> wrote:
...
> Most problematic is the fact that *related* links implies a  
> *relationship*.
> If we are to embrace the concept of a compound location, we have to
> be able to express that relationship.  And that's a whole new can of
> worms.  If Tool A (eg APrompt) expresses a location comprising more
> than one point in the testsubject, how is Tool B (eg Valet) to infer
> anything more meaningful than an _unstructured_ list from that?
> There are a couple of simple cases we can perhaps express:
> * Range (Start-point + End-point)
> * Main+subsidiary locations
> But I think anything more complex has to be tool-specific.
...

Actually, except for maybe a couple of specific cases such as ranges  
(which are actually defined in, for example, one of the  
Xpath/pointer/thingo constellation of specs, IIRC) and perhaps a "happens  
here, should be here too" or something for stuff like missing closing  
tags, I think we should be saying that the list of things has no  
meaningful structure.

I favour using collection in some cases - i.e. this is the complete list,  
but we should note that the order is not significant. I am not sure how  
easy this is to do, since in theory a parseType="Collection" list is  
completely ordered, and it is painful (but not impossible) to construct a  
query that tests only whether it contains the same items.

In other cases we may have a reason to use an open-ended collection like a  
Bag.

By the way, the same structural issue comes up for evidence - that we  
should not imply more meaning to a list of items than where we can develop  
interoperable algorithms and a proof of necessity.

Cheers

Chaals

-- 
Charles McCathieNevile                      chaals@opera.com
          hablo español - je parle français - jeg lærer norsk
         Web dreams are free:   http://www.opera.com/download

Received on Saturday, 24 September 2005 13:27:26 UTC