|
|
Re: [xml-dev] Alternative "character entity" proposal
|
[
Lists Home |
Date Index |
Thread Index
]
Richard Tobin wrote:
Is scoping really necessary?
Yes, it's the main reason that the existing DTD mechanism is
inadequate.
For example, it should be possible to insert some HTML - using the
HTML entities - into a existing document without having to change that
document. MathML is another obvious example.
I think Richard's proposal is pretty good: I have no particular
architectural problems
with it, just a technical druther against adding more scoped
declarations.
But I am not sure that it solves adequately the problems that (I think)
it is predicated on.
All the named reference proposals are predicated that people will be
using tools without
adequate input methods, in particular dumb text editors. (The issue of
inadequate
display methods doesn't seem so important, because the characters that
most people
don't have glyphs for tend not to have strandard entity names either,
except for
the Maths characters.)
Then Richard's proposal provides a way to make your own names. This
naturally
means that, to allow combination, you need scoping to prevent clashing
definitions.
However, if you were to cut some fragment with a text editor, you need
to
bubble down the entity definitions and move them with the fragment (as
is the
case with namespace declarations and ,differently,
@xsi:schemaLocation).
So for this to work well, it really needs some kind of
editor-assistance, otherwise
it is a tedious search. But we have already started from the assumption
we
are using some fairly simple text editor.
But the need for scoping only comes from the need to define your own
entities. Just limiting to the standard entities means there is less
need for
scoped declaration.
I guess there is also a possible intermediate point between my proposal
and Richard's: just the @xmlentfile attribute to allow reference to
a (predefined set?) of known entity declarations, but no @xmlent
attribute. But if people are too lazy (? no...busy) to type in a
dummy DOCTYPE with external references, they would be too
busy for this.
I think Richard's comment suggests another really good point: should
we in fact be moving to a split XML model: a server-side syntax
with numeric character references, entities, CDATA, etc, and
a client-side XML with all sugar free? Servers already do
the XSLT, PHP etc processing: should this be really part of
their job. In other words, should W3C persue a policy of
moving XML for transmission towards Canonical XML, while
simultaneously moving XML for writing towards something
with more sugar (such as ECS, Editing Concrete Syntax[1])?
Cheers
Rick Jelliffe
[1] http://www.topologi.com/resources/pdfs/ECS.pdf
|
|
|
|
|
|