Re: minutes of teleconference, 16 sep

At 7:24 PM 9/16/96, Ron Whitney wrote:
>Notes on HTML-Math Interest Group Teleconference Call
>16 Sep 96

....

> Bruce's first inclination
>might have been to use more of an FA approach, but Dave had made a
>convert of him.

I should also credit Neil Soiffer for this.

>Stephen suggested that this might make extensibility
>more awkward (allowing someone to define an operator with the
>precedence of addition on the left, but of times on the right), and
>Bruce said that, although he felt considerations about extensibility
>are very important issues, they should not necessarily drive the basic
>approach.

I would characterize what I said (or at least, what I wish I said) as:

- the input language, without any extensions, should be as good as
possible with regard to satisfying the diverse design goals for HTML-Math,

- even so, extensibility should be provided,

- Stephen's example is one of the possible extensions that should be
allowed (because it's a member of a class of extensions that has
legitimate uses), even though it would almost always be a bad idea for
anyone to actually extend the syntax in that specific way.


>Bruce and Neil said they felt that users would become familiar with
>the needs of notations they use regularly, and would group expressions
>explicitly otherwise.

I would like to add that, to the extent that we expect authors to
make use of authoring tools, or even just to test their work in
renderers, there will be a way for them to test the parsing of the
expressions they write. In any case, explicit grouping with { }
will suffice if there is any doubt about precedence in author's minds.


>Stephen was concerned about imprecisions which remain in the OPP of
>the Wolfram Proposal, and asked whether a formal grammar existed.
>(There are about 90 different precedences at this point.)  Bruce and
>Neil said they felt one could be written down, although handling
>embellishments on operators complicates matters, and their work for
>Wolfram had actually exceeded a certain internal, hard-coded parameter
>for a public domain YACC.  This is not a limitation in concept,
>however.

We realize that a formal specification of the grammar is necessary,
and I intend to provide one soon, hopefully in time for the meeting.


>The discussion ended with Ron asking Stephen to provide some more
>concrete examples of what might be problematic WP-style encodings
>(i.e. encodings in the Wolfram Proposal for html-math which might be
>ambiguous to readers or might hit the wrong target when rendered to a
>CAS).

Several of the members spoke of the need for such criticism
to be constructive, in the sense that some specific alternative
notation was suggested which, in Stephen's opinion, would be a
better tradeoff, on the whole, for satisfying the diverse design
goals of HTML-Math.

Otherwise it is hard to know what alternatives he has in mind,
or even whether an alternative which avoids the problems he
points out, without introducing worse new ones, actually exists.