Re [office] Processing Instructions

[Jason Harrop <jharrop@speedlegal.com>]

> Jason,
> 
> Jason Harrop wrote:
> 
>> Should our specification state explicitly how compliant applications are
>> to handle processing instructions?
> 
> 
> Well, I wouldn't like that... Please see below.
> 
>> I would expect that by default, an editing application should preserve
>> processing instructions which it does not understand.  (By contrast,
>> when I tested OpenOffice (641 or 643) by manually inserting a processing
>> instruction in the content and then opening it in OpenOffice, editing
>> and saving again, the PI got removed.)
> 
> 
> Indeed. The problem is that processing instructions (PI) are, by and
> large, not part of the content but rather of the physical
> representation. And the physical representation of a document is
> something that we can't (and don't really want to) preserve. The same
> applies to e.g. the text encoding, entities, or indentation.

Text encoding, indentation (and possibly entities, though that can be
argued in more than one way) are not logical properties of a document -
rather, as you say, they are simply a part of the physical
representation (though even then, practical considerations suggest that
some of these are worth quite a lot of effort to preserve - changing the
character set used to write out a document may have no semantic impact,
but can cause all sorts of practical difficulties in various situations).

However, I would argue that processing instructions are _not_ the same
as these. They are part of the logical contents of the document. The
most important part of a document, of course, is the actual text. XML
markup is, however, pretty important, you don't want to lose the actual
structure of your document - markup has meaning. So too do processing
instructions (indeed, PIs are considered to _be_ markup by the XML
spec). Preserving them is crucial to maintaining the document structure.

I understand your point of view, in that the current internal
implementation in OO would make preservation of PIs relatively
difficult. To some degree, this comes down to how we are looking at the
XML document format. Are we buying into XML as a convenient (more or 
less...)
serialisation format for internal word processor state? Or is the intent
to actually adopt XML as it is meant to be used - a document-centric
view of things, where whatever internal state you happen to have should
be exactly that - internal state?

> Jason, a rather basic question: What were you actually trying to achieve? Would other forms of extensibility cover your needs?

Easiest if i show you at the face to face, if we have a minute.

cheers,

Jason