Minutes: TPAC meeting

 Once again, many thanks to Louis for scribing (google wants to correct
this to "scribbling" :-)
Attendees:

   - Neil Soiffer
   - Louis Maher
   - Steve Noble
   - Bert Bos
   - Bruce Miller
   - Sam Dooley
   - David Carlisle
   - Paul Libbrecht
   - Murray Sargent
   - Kevin Ellis
   - John Hudson

Regrets

   - Steven Watt

<https://sandbox.cryptpad.info/code/inner.html?ver=5.1.0-02#cp-md-0-agenda>
Agenda

The attendees introduced themselves.
<https://sandbox.cryptpad.info/code/inner.html?ver=5.1.0-02#cp-md-0-9-00am-general-question-and-answer-for-those-who-do-not-normally-participate-in-mathml-meetings>9:00am:
general question and answer for those who do not normally participate in
MathML meetings

MathML drafts

https://www.w3.org/TR/mathml4/

https://www.w3.org/TR/mathml-core/
<https://sandbox.cryptpad.info/code/inner.html?ver=5.1.0-02#cp-md-0-9-20am-mathml-core-update>9:20am:
MathML core —— update

Discussed "intent to ship" for chrome. It is ready for use except for
printing. The Chromium team is working on print for the new layout engine
that the MathML implementation uses and hopes to finish up in the next
couple of months.

Firefox can already display MathML.

NS: discussed the videos we made for this meeting.

NS: Videos are at https://www.w3.org/2022/09/TPAC/group-updates.html#mathwg
<https://sandbox.cryptpad.info/code/inner.html?ver=5.1.0-02#cp-md-0-9-20am-mathml-amp-media-types-next-steps>9:20am:
MathML & Media Types -- next steps

MathML Media-type Declarations <https://www.w3.org/TR/mathml-media-types/>

PL: Did not receive any comments on his work.
<https://sandbox.cryptpad.info/code/inner.html?ver=5.1.0-02#cp-md-0-9-30am-deep-dive-into-quot-intent-quot-issues-in-an-attempt-to-resolve-them-or-at-least-better-group-them>9:30am:
Deep dive into "intent" issues in an attempt to resolve them or at least
better group them

NS: General intro for those new to the group.

NS: How do you make math speak well? There are different systems to output
speech.

NS: introduced intent with a power example.

NS: discussed prototype methods for speaking things. One prototype is MathCAT
(math Capable Assistive Technology) <https://nsoiffer.github.io/MathCAT/>

PL: discussed combining content MathML with intent.

NS: wanted to show cases of content MathML being converted into
presentation MathML with intent.

DC: suggested that we not discuss content mixed with intent until we find a
case that requires mixing content with intent.

NS: wants to have examples that show content with intent if we are allowing
it in the spec.

SD: offered to show examples of content MathML being converted into
presentation MathML with intent. For most cases, we can start with content
and generate presentation with intent.

PL: sent this: https://w3c.github.io/mathml/#contm_p2s_cn

BM: We are trying to find some content to put intent on. He says we should
not discuss intent with content until we have a compelling need to have it.

NS: Content is not widely used, so finding compelling cases is going to be
difficult.

NS: We are not changing content in this specification, but we could say
that presentation MathML should not contain content.That would mean there
is no need to find examples.

NS: If we do not allow content with presentation, the intent attribute
becomes much simpler.

DC: Mixes content and presentation for his internal work. Shows some
examples from NAG.

DC: In the examples in chapter four we put intent on the presentation and
not in the content.
<https://sandbox.cryptpad.info/code/inner.html?ver=5.1.0-02#cp-md-0-10-30am-break>10:30am:
Break
<https://sandbox.cryptpad.info/code/inner.html?ver=5.1.0-02#cp-md-0-10-45am-continue-deep-dive-into-quot-intent-quot-issues>10:45am:
Continue deep dive into "intent" issues

NS: How do we handle tables. sibling intent references · Issue #402
<https://github.com/w3c/mathml/issues/402>

NS: In TeX, we align things by putting things in tables. How does intent
refer to this without being cumbersome.

DC: There is a problem when the row is longer than one line. You get
wrapped equations.

BM: You should always be able to describe it at the mtable level.

BM: The inability to maintain a complicated structure is why we left out
relative paths.

NS: We will not change the MathML; we will just change how it is read with
intent.

BM: Our speaking rules become very vague to handle complex examples.

NS: How do we bring two rows together. We must use the table approach.

BM: You can devise a rule for speech that works 90 percent of the time. You
will have to come up with a markup for the exceptions which make things
less clean.

BM: You can make an mrow, then make an intent for the entire mrow.

NS: MathCAT has three levels of navigation, the character-based mode, the
simple mode which does not understand structures very well, it does not
know that there is a left hand or right-hand side, and enhanced mode which
does know about structures. The enhanced mode should follow the intent
structure and indicate equation numbers appropriately, gluing the rows
together, etc. It doesn't do this now.

NS: There have been attempts to use overview descriptions to describe
equations, and this was not popular in user studies (ClearSpeak and
Karshmer's editor). Nonetheless, we both believe that it should be useful.
Not enough effort/thought was put into MathPlayer's implementation and that
might be why it wasn't popular.

BM: It is a good idea to be told the shape of the system of equations
before you describe them in detail.

NS: You can put a line break in a long equation row.

BM: We can work with what we have now. (DC showed examples of diagonal
matrices, lower-triangular matrices, and matrices with sub-matrices. How do
you speak those? The interesting part may be that it is a triangular
matrix.)

NS: His guide is to follow how things are spoken in a classroom, although
for some things, it isn't clear how they are spoken in the classroom.

All: some discussion of special matrix forms such as tri-diagonal matrices
and block matrices.

PL: says that the internationalization group wanted to have a survey of how
things are spoken in classrooms for various languages.

PL: We did that for long division in the past, with interesting results.

NS: Speaking the long division, the trick is recognizing is that it shows a
process, step1, step2, etc. MathPlayer describes a step, says "next step",
and repeats the pattern.

PL: So, survey is useful. Might be good to not just get text, but also
record the voice.

PL: Suggested that we record people speaking mathematics in various
languages.

NS: We would not understand what they are saying. Who would understand
Chinese?

**ACTION* PL: will start considering how we can survey how math is spoken
in classrooms using other languages.

BM: If you read everything, you may have to put intent on each entity.

DG: proposed that as you read each number, the number type would be
described. You would be told if the number were real or complex.

NS: Does overviews of expressions and allows you to zoom down into the
various expressions.

NS: We have a list of intents some in core and some open. We must decide
what to do with all these intents. That will be a topic for another day.

NS: Summary: we will live with ugliness for tables.

BM: suggested looking at n-ary Intent discussion · Issue #253
<https://github.com/w3c/mathml/issues/253> to eliminate some ugliness.

DC: began a discussion on how "*"s could be used in intent to refer to a
list of arguments.

NS: thanked everyone for sticking around for the long meeting.