[
Lists Home |
Date Index |
Thread Index
]
Hi David,
>> John Cowan and I have been pushing around the idea that it should
>> be a requirement on 'official' infoset extensions that they come
>> with a serialisation. I'd like to see this happen, for example
>> along the lines of the various reflections Richard Tobin and I have
>> explored for the PSVI. It would be a simple step to allow XPaths to
>> navigate extensions via a pretend reflection. All that's required
>> is a function 'reflect(property)', whose value would be a
>> reflection of the value of an infoset property of the current node.
>> So, for instance, to access the appinfo annotation of the type used
>> to validate an 'apple' element child of the current node, you'd
>> say:
>>
>> apple/reflect("type_definition")/annotation/appinfo
>>
>> You could of course navigate the vanilla infoset this way. The
>> equivalent of /*/apple/@banana would look like
>>
>> /property("document_element")/children/element[local_name="apple"]/attributes/attribute[local_name="banana"]
>
> Yes, direct PSVI access would be one possible application of an
> annotation axis or function (as Jeni suggested with
> foo/#psvi:type-definition in an earlier post), but I see no point in
> limiting it to PSVI and PSVI only access as your reflect function
> suggests.
I don't think that Henry was suggesting limiting it to PSVI -- he was
just using examples from the PSVI (actually, they're from the basic
Infoset, I think) because it saves him from making up anything.
I think that the basic point he makes -- that whenever someone adds
something new to the Infoset they should also provide a standard XML
representation of that -- is a very good one that would really support
the idea of accessing annotations as a node set using an axis or
function. Very cool.
Cheers,
Jeni
---
Jeni Tennison
http://www.jenitennison.com/
|
