Re: A few notes on text

Claude L. Bullard ([email protected])
Tue, 6 Jun 1995 10:30:36 -0400


[Mitra]

| This looks like re-inventing the wheel, all the same discussions that
| happened over in HTML are going to happen again in VRML.
| Its a pity really that we didn't just go with HTML for the text, then we
| wouldn't have to re-invent the wheel again here. Even if full HTML was too
| difficult for most browser writers, we could have at least used a subset.
| The current spec looks like a re-invention of the early days of
| typesetting.

One may note that those discussions occurred because those
who thought to create an international WWW tripped
ever so badly when the Western programming culture
discovered how complex international text representation is.

Gavin presents one of the *simplest* approaches uncovered in
those discussions. If VRML can use only its graphic
primitives for text, VRML text will not easily
be available to other systems that want to reuse it.
Using the VRML primitives is "reinventing the wheel".
Using HTML attaches the wheel to a broken axle.

HTML IS a *typesetting* solution as it is
a presentation application. If one elects to use
HTML directly, one inherits that. The current shootout
between HTML and PDF indicates the awkward
aspects of this. BTW, the early days of typesetting
looked very little like current HTML. The middle days
(somewhere between Unix utilites and DS-Runoff) looked
like HTML, and then only superficially.

There is a great deal of work going on in the area
of internationalization of text sources. Gavin is
taking advantage of it and his extensive experience
with non-Western cultures and languages. If VRML
can't get this one right, it is one more reason not
to use VRML. VR needs text whether VRML does or not.
Unicode is a good default option.

Chris Marrin's comments are cogent. Making family
an MFString appears reasonable *on the face of it*. |-)
I disagree that supporting only one encoding is prudent.
Choose a default, but leave the option open. Some
options aren't "emotional"; they support legacies
or they express intent. Failure to enable *intensional*
support is sufficient reason to reject
VRML-based systems; so, I think it imprudent
to 'hogtie' system vendors. Choosing
a default, OTH, is a perfectly acceptable
expression of the intent of the VRML designers
and sets the precedent. Unification then
occurs by consent.

Len Bullard