Hi all,

Some of us had previously discussed the possibility of taking MathML 
seriously and enabling users who prefer it to skip KaTeX altogether. The 
built-in XML support makes it possible to input MathML, but there have 
been some obstacles to this being practicable (in addition to the 
verbosity, which persists to some degree even if you wrap things in 
macros).  I have a feeling the verbosity can be solved and am actively 
thinking about it, but the main show-stopping obstacle has rather been 
the integration with LaTeX for rendering diagrams (e.g. tikz stuff).

Currently, the “TeXlike” content sent to KaTeX can equally well be 
sent to tikz, and so things mostly just work. But if MathML was the base 
format, there would be no TeXlike content to send to tikz; I thought for 
some time about whether we could do things like convert the TeXlike code 
to LaTeX (e.g. by doing an XSL transform on the fly) but I think this is 
ultimately a bad idea, not to mention one that would be very hard to 
implement in the Forester tool.

I realised something that could give us a way out: Context actually has 
built-in support for MathML, and you can even dump MathML blobs into 
ConTeXt code. ConTeXt supports tikz as well (though you have to be a 
little careful with the syntax). But here’s an example of some ConTeXt 
code that renders properly (with context --luatex):

	\usemodule[mathml]
	\usemodule[tikz]
	\usetikzlibrary[cd]

	\starttext

	\starttikzcd
	\XMLdata{
	 <m:math xmlns:m='http://www.w3c.org/mathml'>
	  <m:mrow>
	   <m:msup>
	    <m:mi>x</m:mi>
	    <m:mn>2</m:mn>
	   </m:msup>
	  </m:mrow>
	 </m:math>
	}
	\arrow[r]
	\NC
	\XMLdata{
	 <m:math xmlns:m='http://www.w3c.org/mathml'>
	  <m:mi>y</m:mi>
	 </m:math>
	}
	\stoptikzcd

	\stoptext


I think that this is viable… Many users will prefer to use KaTeX + 
LaTeX, but there is an amazing variety of more advanced kinds of content 
we can support through direct use of MathML, and I think that advanced 
users like myself would probably like to use it if we can make it 
friendly enough. Given that the above resolves the main technical 
problem, I think it is possible to start brainstorming how to facilitate 
usable macro libraries for generating MathML code. I can imagine a 
couple features that would be really useful, for example:

+ A way to take an XML attribute list as an argument in a macro and pass 
it to a given XML element.
+ A feature that lets you write macros that automatically enclose 
themselves in some XML element (e.g. to be able to write `\mi{foo}` 
instead of `\math{\mi{foo}}` in various places).

Best,
Jon