addressing DBooth3

So, if we want to go to PR, we need to agree a response to comment DBooth3.

We can agree to design changes, such as David is suggesting, and take 
some more time to make sure that:
- these changes are motivated by use cases that we understand
- we have code that demonstrates the features of these designs
- the new code does address the use cases,
- the possibility of unfortunate interactions with other specs that pass 
infosets or nodesets around, and do not have the bytes of a document to 
hand, has been appropriately reviewed
i.e. that the new design works.

Or we can agree that the current design is OK; and respond to David 
indicating the changes we have already made in response to his comment.

We may want to offer David (and the world) some rationale for the 
current design.
At the heart is the #faithful-infoset issue.
While I was not present during the discussions, there seems to me to be 
very good rationale behind it.
=====

The Web works by mapping sequences of bytes to representations of 
information.

There is variability in these mappings.
This variability is part of the contract between producer and consumer 
discussed in my previous message.

Along with desirable variability, there is also some unfortunate 
variability, that is largely historical. It is also not clear that the 
current balance between producer-and-consumer is always appropriate. 
That is why, as long as there is the Web, there will be issues like the 
tag issue xmlFunctions-34, as part of the management of that variability.

The GRDDL WG is not charted to address that variability.

Like other WGs, the GRDDL group has worked within that space of variable 
mappings, and come up with a key rule of thumb for ensuring that the 
variability is not a show-stopper:

- the publisher is responsible that all GRDDL results are not untrue - 
noting that the main problem is producing incorrect triples, not 
producing too few.

Like many other specifications, the GRDDL spec, builds on top of an 
abstraction of a document, the XPath Nodeset. This ensures that the 
GRDDL spec will work in situations where there isn't a document to hand, 
and can operate within some other application process.

====

Jeremy


-- 
Hewlett-Packard Limited
registered Office: Cain Road, Bracknell, Berks RG12 1HN
Registered No: 690597 England

Received on Friday, 22 June 2007 09:10:48 UTC