notes on conference call of 19 august

Notes on HTML-Math ERB Conference Call
19 August 96

In attendance:

Patrick Ion	Mathematical Reviews
Bruce Smith	Wolfram Research, Inc.
Neil Soiffer	Wolfram Research, Inc.
Bob Sutor	IBM, Yorktown
Ron Whitney     American Mathematical Society
Ka-Ping Yee	Alias|Wavefront, Inc.

[Notes prepared by RW.  Corrections welcome.]

Ron opened by discussing whether and what to send of our materials to
Safir for review.  Neil, in off-line conversation, had indicated
concern that the Wolfram proposal's incomplete state might lead to
wasted time on the part of reviewers.  The teleconferencing group felt
that, given Stephen Watt's current membership in the larger ERB, Safir
could actually access whatever materials it wished and review
accordingly.  For the Wolfram proposal, Bruce suggested that Safir
start with the posting of 31 May and read later notes and
correspondence for updates and modifications.  Correspondence and
position papers posted earlier than 31 May run the risk of being quite
out of date.  All materials we are currently considering are
accessible from the AMS ERB page as starting point (this includes both
the Wolfram proposal and a link to the MINSE pages).

Conversation then shifted to a discussion of the appropriateness of
ISO 12083 as a starting point for Wolfram's "display list"
specification.  Bruce pointed out that display list format was not
intended as only a point of departure for visual rendering, but as the
base for all renderings (visual, audio, CAS, etc.).  Bruce (in the
Wolfram proposal of 31 May) described display list format as a stage
of processing so as to delineate where template-matching would occur
in the sequence of events (to wit, template matching operates on
the expression tree and is part of the process which creates the
display list).

Bruce felt that ISO 12083 is *not* a good base for display list format
since 12083 need not contain the full "expression" information which
display list format does (e.g. the scripting examples posted by RW
show that the "base" of an object scripted needn't be marked as such).
This is also a point which Neil had made in earlier conversations.
Bruce felt that, for those using HTML-Math to present legacy data on
the Web, the options are to either "upgrade" to HTML-Math by filtering
with a local operator-precedence scheme or to accept the possible
negative formatting consequences of displaying with "degraded"
information.  Ron said that he could forsee AMS production
accommodating an operator precedence scheme within its production
stream in the future, but didn't think legacy data would be filtered
using one.

In any case, since the intent of display list format is not to produce
an intermediate "visual display" data structure, the view that ISO
12083 can be used for such a thing requires modification.  Visual
browsing will be part of the software we present to W3C as a model for
HTML-Math.  Whatever specification we give for HTML-Math itself,
browsers (or plug-ins) will have to pass to visual display.  Whether a
canonical display language (such ISO 12083 or TeX) can interface
cleanly to this remains open.  It is a stated goal of our committee
that filtering from visually oriented markup to HTML-Math be possible,
but the quality of the result and the requirements for filtering have
been left open.

Discussion then moved toward semantics and modes of capturing
semantics in the Wolfram proposal and MINSE.  Neil said he felt that
the issue of semantics was very large and ill-defined (everyone's
notion of semantics is different).  Patrick acknowledged this, but
also commented that attendees at a Panel Discussion at the summer AMS
meeting last week in Seattle showed great interest in Internet
communication on the K-12 level.  Since K-12 semantics is much more
straightforward, we shouldn't discard (not that Neil said he wanted
to) the possibilities of enabling robust semantical markup.  Bob
pointed out that, at least initially, most semantical markup may be
provided by software (CAS for example) and not human operators.  Ping
remarked that he felt MINSE allowed a good entry to semantics and that
it enables downloading style and notation definitions.

Bob asked whether the idea of textual annotation has been considered
in our deliberations.  We have indeed discussed this to some extent,
although problems have not been entirely solved.  There is facility
for adding text strings in both the Wolfram and MINSE approaches,
although we have not seen our way through the full problem of text
within math within text within ... .

The phone call ended with exhortations by Ping to comment on MINSE
features, and by Ron to provide concrete agenda for the Sep30/Oct1