Minutes: MathML semantics meeting, Sep 3, 2020

Thanks to Brian for helping with the minutes.

The meeting was recorded:
https://benetech.zoom.us/rec/share/BoKJT1IgTxEl4IREcSG4EwU7VWPTA2_5D8qYKHF5eZX610OZyNibKjMoBtEZGs-a.z_I0ZfPYHyxx5ofI
Passcode: Oy+8?$mF (starts around 8min in)


Attendees:

Neil Soiffer

Deyan Ginev

Louis Maher

Murray Sargent

Bruce Miller

Moritz Schubotz

Sam Dooley

Brian Kardell

David Farmer

Patrick Ion

Regrets: David Carlise, Steve Noble


Charter is at
https://docs.google.com/document/d/1W-oYUbOMueaqb3KFSWkjWVBwR6AzSEBizHwQhvSwfDc/edit#

Updates

DG: I’ve finished the letter ‘C’ for level three naming. That’s ~250 names.

BK: The TAG has a debate about semantics that is ongoing. There are
discussions about semantics. <p> was introduced mainly for spacing reasons,
but it is semantics. Accessibility is addressed in the AAM and that’s about
semantics, HTML is not about semantics.

BK: Cautionary tale: elements were added to HTML after years of debate and
things were used more often things that weren’t intended.

SD: So is the meaning of what something is what someone uses it for?

NS: Things like <article> were introduced. Are those not used appropriately.

BK: HTML5 tried to add <section> but it isn’t really used. It was supposed
to replace <h1>, ….

DG: Are you saying we should use “data-”?

BK: I’m not suggesting anything. It took about 10 years to get <main> into
HTML. Wasn’t worth the effort.

BK: If it isn’t in the AAM, it isn’t useful.

NS: AFAIK, AAM doesn’t exist as a rec.

BK: It's definitely real, formal, exists-  it's linked up… HTML's:
https://www.w3.org/TR/html-aam-1.0/
      SVGS: https://www.w3.org/TR/svg-aam-1.0/  and MathML's has begun, I
believe it is not currently official yet but it is started in
https://w3c.github.io/mathml-aam/

1. Semantics names -- continue discussion   a) Should the names brought
over from Content MathML be changed to follow this naming convention? [I
don't think this was resolved]

DG: I would recommend we follow the general guidelines and have the short
names (e.g., “lt”) as an alias.

General agreement, modulo what “alias” means.
   b) More discussion on the meaning/purpose of the various levels

NS: DG, can you summarize from last time?

DG: The way I saw it was a robust agreement.  It's about drawing a
distinction between the levels and what we are investing time into as a
group.  Last week we kind of reached this ? about what symbols mean.  The
way I have been thinking about the levels which is not final, it's just my
thoughts.  The first level - level 1 is what you need.  Level 3 is kind of
a scratchpad. I have a lot of annotations that are nice for some purposes,
but maybe not necessary.  Level 2 is kind of the mechanism where things
move from level 3, into a state before level 1.

NS: Along these lines, I see a parallel to Feyman in the 60s.  It got to be
popularized and now it might be popular enough that it moves into L1 at
some point.

BK: I like this transition from level 3 to level 1. How do you imagine
collecting the data to know whether it should move between levels?

DG: I like the idea of growing the community to who can add to it, we just
give access to level 2.  Level 3 is wide open.

BK: I think we need something like this, but I don’t know of examples. The
closest thing is info gathering on usage. I don’t think it is a great way
to collect info on math though. I think it would be good to formalize how
you do the move.

PI: I don’t necessarily agree. It isn’t Unicode or AFII. I think the people
around MathML know a lot about what is trending and important. We need to
get something going that is going to use these features.

BK: In 30 years, HTML was able to standardize 130 elements. ARIA even less.
Those are also "small things"
NS: The core set of things are so common that people have special ways of
notating them or speaking them. For example, fractions are spoken
differently depending on the content three eighths or 3 over 8 or ... Those
need to be given to AT so that AT knows it should employ the special rules
based on content, or on the user. As you move away from those, things
become more regular and don’t need specification of what to do, you just
need to pass the info to AT and AT or search takes it verbatim and doesn’t
need to interpret it to do something special, it can just speak it. “Plus”
is another special case where search may want to do something special
because addition is commutative.

DG: That’s what I think is right, just standardize a few things and leave
the rest open-ended. Level 3 is useful for many things.  One of the big
challenges with content mathml is that no one wants to write a content
dictionary.

BM: I think there is something that is not well understood about MathML as
it currently exists…  There is a set of names but csymbols are open-ended
so people could add anything that want. But the problem is the OpenMath
community had problems standardizing those symbols. There needs to be a
community that deals with this list. All of the symbols should not be part
of the standard, it should be open-ended. But we need to be cognizant of
problems of the past.

PI: This isn’t that different from Unicode where all the symbols needed to
be defined. They came out of the TeX of the community but the names were
not that useful. It wasn’t a big community activity, it was a smaller group.

DG: That’s a good example. It is a fixed list that can be searched. As
opposed to OpenMath where adding something is hard.

BM: it would be best to focus on the essential terms.

MS: For what it is worth, Unicode 9.0 has 3,310 glyphs that have the math
property.

NS: I just finished going through a bunch of them and it's not clear what
many of them are used/useful for

MS: I would be delighted to include things that enhance the understanding
of this.

NS: Pointers off to examples of use, uses, etc would be great.  I think the
process now is better, early on there were less requirements of these sorts.

MS: Some of these were given to us by groups that have said "we have been
using these hundreds of symbols for some time now" and there aren't good
references to where or how.

NS: Ok, let's talk about aliases… I personally want just one name to look up

DG: lt is smaller than less than

NS: I could easily imagine almost everything in core having between 3 and
10 aliases - is that actually beneficial?

PI: Less than or equal is often used for normal subgroup

DG: I see it as easily technically resolved, I think it is fine to have 1
for L0.  For L3, it would be impossible to have just 1 name.

BK: Maybe with math, it is less this way. There is a primary

NS: It's more about identifying what it is and having some real list than a
name…

BM: To the extent we have only one term for every concept in level 0, then
there is not any reason not to use the content MathML names. In level 2 and
3, there we want hyphenated names.

DG: I think it is an optics problem, because it would be different from 2
and 3.

BM: But level 0 is different. But I’m not advocating one way or the other.
I’m just saying if it is small, I don’t see the need for consistency.

NS: I think there is a potential optics problem the other way if it says we
are replacing content MathML.