Re: Consider removing the MathML entities from text/html

On Fri, 02 Oct 2009 15:29:49 +0200, David Carlisle <davidc@nag.co.uk>  
wrote:

>
> Simon Pieters wrote
>>
>>
>> I'd like to question this decision.
>>
>>    * MathML entities' reason for existence seem to be based on the
>>      assumption that people will write MathML markup by hand. If it's  
>> not
>>      written by hand, it makes no sense to have entities. Given the
>>      verboseness of MathML, I would be surprised to find anyone to write
>>      MathML markup by hand more than as a try-once-never-again  
>> experience.
>
> Actually given a little editor help for closing elements, mathml isn't a
> lot harder to author by hand than other markup methods for mathematics
> (which are all far more markup intensive than conventional text) It's
> not at all uncommon for people to hand author mathml fragments as part
> of larger works.

I'm surprised.


> Also for most people, even if the mathml is converted
> from, say, latex, If you need to modiufy or check the markup, it is
> easier to recognise the symbol in entity form than as a numeric
> reference, or as character data (this last may change if editing
> software picks up the ability to use math symbol fonts)

Yeah.


>>    * Have the MathML entities been researched for compat with existing
>>      text/html content? For instance, do pages expect "&wp;" to show  
>> those
>>      four characters rather than showing "%1B%G℘%1B%@"? I think it  
>> would be good to
>>      research this.
>
> pass
>
>>
>>    * Do any commonly-used MathML-emitting tools emit entities? Microsoft
>>      Word apparently doesn't. I think it would be good to research
>>    this.
>
> One very common piece of software that always (I think) shows mathml
> characters as entities, even if the source does not use them is
> firefox's "view mathml source" right menu option, so anyone taking
> fragments from there will currently get a fragment using entity  
> references.

Ah. I didn't know Firefox's view MathML source feature serialized with  
entities. That doesn't work with pasting into XML without the right  
doctype. Maybe XML should have all the entities as predefined?

I also noticed that Firefox's view MathML source doesn't serialize the  
namespace declaration if there was no xmlns attribute in the DOM, which  
also doesn't work when pasting into XML.


>>    * Having 2000+ entities instead of 200+ entities doesn't seem too  
>> nice
>>      for implementors.
>>
>
> There is obviously a cost and I can't speak about browser
> implementations but I'd have thought that overall it's better for users
> to have them always there, built in, rather than as now, only having
> them if the browser chooses to read a dtd and even if as in firefox's
> case that is a local dtd in its installation area, still an appreciable
> cost and no cross platform guarantee that they will work at all.

Indeed.

-- 
Simon Pieters
Opera Software

Received on Monday, 5 October 2009 10:18:06 UTC