[
Lists Home |
Date Index |
Thread Index
]
Tim Bray wrote:
> Bill de hÓra wrote:
>
>> Even so, I'd like some words around whitespace - it's great for
>> thrashing out details. And I was thinking of cases like
>>
>> <foo& ;bar="baz"& ;/>
>
> This clearly not well-formed XML and is the same in UTF-8 and +names.
> The issue is whether +names needs to declare any additional rules about
> what can come between the & and ;. At the moment I think not; since
> there's an exhaustive list of replacements, and a statement that
> anything else is just passed through, I think that covers it.
Ok, but maybe say that in the next draft, even using & ; as an
example - it wasn't obvious to me that & ; is not ok.
> There is a well-known problem with XML, namely XML documents tend to not
> nest nicely inside other XML documents. I think we're stuck with that
> one. It's totally the case that if you're going to use something like
> +names, you own the responsibility for ensuring that anyone to whom you
> send gets told firmly what it is. The encoding declaration helps a lot,
> but there are going to be cases where breakage will occur unless you put
> in some extra effort.
No doubt. Again, I think you could take what you've just said out of
the first person and use it as words for the spec. Thre's lots to be
said for an I-D that articulates any caveats or risks.
> So far we haven't really heard very much from the
> constituencies whose issues this is designed to address. If this
> silence continues, the conclusion will be obvious.
Erm, you can count me in the that group. I've needed to whizz
entities about the network, I've needed to deal with cutnpaste, and
I'll need to do so in the future.
Concerns in a nutshell:
o roundtripping. I'm hosed unless every XML-aware process in the
chain is +names aware (unless I have a pipes and filters arch to
transcode the names to numerics). I think if this draft gets
traction here and elsewhere in client tools, we'll need to push very
hard to get it adopted on the server also.
o upgrade path. When I need more entities, what's the procedure
for encoding them? If the toolbuilders were smart about how they
code this (again pipes and filters) maybe we could declare new
encodings a la mimetypes (tho' not sure that's a good road to follow).
Bill de hÓra
--
Technical Architect
Propylon
http://www.propylon.com
|
