Re: [CDR Framework] i18n comment: Language identification for child documents

* François Yergeau wrote:
>xml:lang is explicit language information.  The issue is that the CDR 
>spec should say that its scope extends to referred-to objects in a 
>compound document, just like it deals with events, so that the CDR works 
>just like a CDI.

It seems to deal with events rather poorly, the proposed model is quite
incomplete (events just bubble from the child to the parent, they do not
accomplish the capture phase in the parent, for example) and insecure.
In some environments it might not even be implementable this way. Also
note that CDR even with this model is still very different from CDI,
e.g. styles do not inherit across the reference boundaries as far as I
can tell from the draft.

>Did you read the little scenario I made up?  Don't you think that *not* 
>changing the styling/behaviour when going from a CDI to an equivalent 
>CDR is better that changing it to some random outcome based on guessing?

Well, your scenario seemed rather silly, if you don't copy the language
information over it gets lost. If you don't want that, don't lose the
language information. You could make the same point about scripts that
break if you take the SVG fragment out of the document, they break since
they don't access the child content through the ReferencedDocument et
al. features. The solution for CDR-like-CDI is XInclude with a proper
security model.

>I'm afraid that specs don't define good results.

That's why I wondered why your model gives better results than others.
Björn Höhrmann · ·
Weinh. Str. 22 · Telefon: +49(0)621/4309674 ·
68309 Mannheim · PGP Pub. KeyID: 0xA4357E78 · 

Received on Friday, 10 February 2006 16:34:24 UTC