[
Lists Home |
Date Index |
Thread Index
]
Rick Jelliffe wrote:
> ??? I believe Bill's objection is that the recipients will (always) rationally choose what
> information or even tags they consider of interest. No objections.
>
> However, in the absense of special knowledge, they cannot know what was in the external
> subset, so they cannot rationally choose. They can only hope. An instance in which
> declarations are arbitrarily not received is a corrupt instance.
I neither missed nor ignored this point in your post, nor do I believe that I misunderstand nor
fail to comprehend it. However, to judge by your very words ("For data/document transfer, there
should be no variability of the infoset possible: what you send is what they get. I think it
is a basic matter of data integrity.") you are asserting something far more momentous than the
'packaging problem', for that is nothing more than the logistics for assuring that what the
document sender understands as the complete instance arrives with all of its parts at least
available to the recipient. You speak of the variability of the infoset. Grant, as I gladly do,
that proper document handling needs to assure that the whole of a document instance is
available to each user and you still will not insure the invariability of the infoset. The
nature of XML, even drastically subsetted, demands the parsing of a text instance at every use
of an XML document. An infoset, if it is anything at all, is built on the output of each such
parse. It is not the textual instance input to the parse. It is not, except in some fevered
imaginations, the platonic form nor fluid preverbal Gestalt of that instance. An infoset is
separately elaborated from the operation of process on the occasion of each parse. Assuring "no
variability" of such semantics will require more than assuring the integrity in transmission of
the instance document as its creator understands it. No variability in the outcome of process
will require no variability in the processes themselves, regardless of their local purpose or
local requirements. That requirement will quickly become the requirement of a single canonical
process and thereby of a single orthodox, 'intended' use of each document. You and I, Rick,
have both heard those who ask for precisely that. You know better than to allow the text to be
chained to that pulpit.
|
