Re: Relationship types
Joe English wrote:
> [***] My understanding of "the MID model" is based on the
> November 1994 draft spec; I've not yet read up on
> the current version, but it appears to have some
> new functionality, including a multi-anchor, general-purpose
> "relationship" link that would probably work as
> an architecural base for "CLASSLINK".
Right. The gosub/goto/spawn/other stuff is there in case
the author thinks they need it. It is necessary to look
at the original MID requirements in some detail to understand
what the sponsor was after. I'm not sure those are available.
I was quite surprised when I went to the navycals site to
discover how much of the original MID documentation is missing
or altered. That is very disappointing because without that
documentation, it is difficult to understand the reasoning
behind what we did. I may have some of that still in my
possession, and may at some point in the future put it up
on a site so it can be read. It is important to understand
that the sponsor wanted *behavioral specs* and *look and feel*
in the document type. We almost ended up with a CGM approach.
I have been studying the TEI extended pointers. It includes
a type attribute which appears to be used when the author
wants to declare something (e.g, type=navigator). It
also appears this could be used should the author decides
to include state management hints. Lou's document is
quite clear and easy to grasp. This is BabyHyTime and
something like that is a very good place to start with XML 1.0.
I think it should be expressed as an architecture, but it might be just
as easy to make a parameter entity and get the same effect.
So which chicken or egg came first: reftype or targtype?
Just reading this, it looks like the authors went out
of their way to use HyTime concepts but change the
names to protect something or another. Reftype has
been around since Caps asked for it to use in 87269.
I'm still not sure how one would specify chained traversal
to prevent goto/label where not safe, but I've not had
the document for very long. What I wanted in MID was a
sequence of locations one could point to using a nameloc
and implying that identifier order was strict, but I
was overruled on that one. I thought it easier to
say, "here is a list of id/locations. Traverse in
sequence". So we used the chain element type and
containment instead of indirection.
Regardless of how folks like it, scripts in declarative
markup are now a way of life and will not go away. As long
as something like a script node can be declared by XML
applications who wish to use the approach, fine. I do
not subscribe to the SGML Way on this issue. Too hard
on performance, and frankly, silly. Nothing in SGML
except Charles' promise says it can't be used to
create a scripting language, and in fact, it has
been done several times. Slow ones, to be sure.
In any event, a normative list is not needed. Examples
will do just fine.