Re: Ideas for future improvements

Le 05/12/2014 09:05, William F Hammond a écrit :
> Not if the fences wind up being constructed with CSS as enhanced
> border decorations (somewhat beyond what is presently available in
> CSS) with precise effortless stretching as a bonus. Having the fence
> specs in <mo>'s would be awkward for that.
This won't work for the "separator" attribute or for general Unicode
characters that don't have a CSS border definition. And we still need to
implement stretchy rules anyway, so using CSS is only helpful for the
simplest cases like those of Opera's stylesheet.
> Moreover, the fence operators are not semantically the same as other
> operators, and there is no definitive way to tell whether operators
> that are fence-like first and last children in <mrow> are really there
> as fences. For example, in [f(x)]_{x=a}^{x=b} (which resolves as math
> to f(b)-f(a)) the brackets are non-fence operators. Of course,
> presentation mathml is semantically poor from the outset, but this
> doesn't mean that whatever semantics are there should be trashed. 
Then I believe you are doing something wrong. The spec says that "any
|mfenced| element is completely equivalent to an expanded form described
below" and "In particular, authors cannot be guaranteed that MathML
preprocessors won't replace occurrences of mfenced with equivalent
expanded forms. ". So if you want to preserve any kind of semantics you
should use a "class" attribute or a <semantics> element with parallel
markup and not rely on a presentation element name.
> Don't forget that there are many implementations other than browser
> rendering, both for generating mathml and for consuming it. If people
> believe that MathML documents are not durable at least for the extent
> of their lifetimes, they will run from MathML. [...] It would also
> break the paradigm of XML. A document spec-ed as XHTML 1 + MathML 2
> should be durable forever. [...]  BTW, I believe there has been this
> debate about <mfenced> virtually from the beginning of mathml.
What you say makes sense, but it's probably this kind of reasoning that
makes MathML not attractive to Web browser companies.

Frédéric Wang

Received on Friday, 5 December 2014 10:30:28 UTC