Minutes: MathML Full meeting 4 Sept, 2025 (Zoom summary)

Here is a lightly edited zoom summary for the 4 Sept meeting for those who
want more information...

*Summary*

The team discussed ongoing work at the DocEng conference in Nottingham and
received positive feedback on their LaTeX implementations, while also
addressing technical challenges with PDF accessibility and MathML rendering
issues. They also reviewed unresolved issues and discussed potential new
properties for
consistent reading of fractions. The conversation ended with discussions
about
polyfills, algorithm documentation, and language support capabilities for
Mathcat, including file packaging and compression methods.

*DocEng Conference MathML Presentations*

David reported that the DocEng conference in Nottingham is currently
underway,
with several team members presenting a workshop and talk on MathML in PDF.
The
team has received positive feedback on their LaTeX work, which was
surprising
to some attendees as it had not worked properly in the past. Neil and David
discussed ongoing technical issues with PDF accessibility, including
problems
with NVDA and Foxit, which they are investigating but have not yet resolved.

*Accessibility Project AI Integration*

The team discussed the challenges and future directions of their
accessibility
project, focusing on the integration of AI and rule-based systems like
MathCat. They explored how AI might replace MathCat's speech functions
while
maintaining braille capabilities, and considered the potential for AI to
assist with intent annotations. Moritz inquired about the stability of the
Mathematical Core specification, which is currently a candidate
recommendation, and Neil explained that this stage invites implementers to
test and provide feedback. The team also addressed limitations in linking
functionality within the Core specification, noting that while href
attributes
are not supported in Chrome, workarounds such as JavaScript polyfills or
using
A elements are available.

*Fraction Reading Property Discussion*

The team discussed adding an ordinal or "over" property for consistent
reading
of fractions, with Neil proposing it as either a property or core concept.
David and Deyan agreed it should be a property rather than a core concept,
as
it's too specific. Neil explained the history of ClearSpeak, a standardized
reading system for accessibility testing, and mentioned that some of its
rules
could be implemented as properties. The conversation ended with a decision
that Neil would propose a list of properties based on the Mascat system, as
they currently don't have enough properties to properly implement this
feature.

*MathML Polyfills Issue Resolution*

The team discussed closing an old MathML issue related to polyfills, noting
that the polyfills had been implemented and tested manually. David
mentioned
he had created a rolled-up version of the polyfills but hadn't pushed it to
a
branch yet. Paul raised concerns about the packaging of polyfills,
suggesting
they needed to review which ones to include and which methods to expose.
The
team agreed to close the issue, with David planning to push his polyfill
work
to a branch for review. Neil then brought up the remaining substantial
MathML
issue about ordering and references, which had been a topic of ongoing
discussion.

*Algorithm Documentation Simplification Discussion (#449)*

The team discussed how to specify algorithm behavior in documentation, with
David and Neil expressing concerns about making the specification too
complex,
while Bruce and Murray emphasized the need for clear expectations in edge
cases. Deyan clarified that the algorithm code was intended as an example
rather than a mandatory part of the specification. The group agreed that
while
the algorithm might be complex for specific cases, most practical
applications
would be straightforward, and they decided to focus on making the concept
simple for authors to understand rather than getting caught up in
implementation details. Neil noted that Deyan had been tasked with writing
a
comparison between David's statement and the existing implementation, which
was still pending.