Re: 'Newbie' questions

[Steve Witham <sw@xxxxxxxx>]

Aaron Bingham asks:

> * The Udanax Green back end can supposedly store any type of data, but
>   how can the front end determine the type of a document, since the
>   back end does not seem to store any type information?

These are somewhat-amateur answers.  A lot of the ideas can be gleaned
from the book Literary Machines, which is a more user-level view of
Xanadu.  Xanadu has a very VERY general, in other words fairly open-
ended, way of specifying types.  Types exist on the border between
guts of how the system works, and conventions of how it is to be
used by front ends.

Types are done with links, either on documents as a whole, or on
parts of the documents.  A link has parts (I forget the order):
    < fromset, type [, toset] >   (or is it the *from* that's optional?)
Also, the link lives in a location, which can be anywhere including
inside the from document.  I mean the link is part of some document,
possibly the document it's about, possibly another document.
You would imagine that typically a document includes a link about
its own type.

The fromset is a set of ranges of addresses.  A document-as-a-whole
has one address and its contents are another range of addresses.
So, a link from a document-as-a-whole with a type asserts that the
document has that type.  A link from some of the contents with a
type asserts type info on those parts of the contents.

The idea was to have some places in the Xanadu address space be
types.  I believe some people (Mike Butler?) worked on type systems.
This was before MIME, I think.  Also before XML--if you know XML,
it specifies a document type as a link to a place that defines the
type.

But, I don't know that an ironclad, totally-agreed-upon setup
for the exact system of document and content types exists.
It would seem smart to hijack MIME for document types.

It would be nice to have places for the type systems of RTF,
HTML and XML mapped in somewhere.

This raises the issue: it would be nice to have a mapping from
URL space to Xanadu space.  Or, for that matter, a way to start
carving up Xanadu space.  (Xanadu addresses are like IP version
infinity, with everyone a sub-authority, but still someone has
to be the top.)

> * How would multi-media documents--here meaning those containing any
>   two distinct types of data, such as text and graphics--be
>   implemented?  Must they be constructed from separate documents (in
>   the Green sense) using transclusions, or is there another mechanism?

Documents have tumbler (that is, hierarchically) addressed innards.
The major division in the content of every document is between the,
I don't know the term, the body, versus the links.  But inside the
body, hierarchy is possible too. You can put type links on the different
parts.

Here, a mapping to MIME multipart would be nice.

> P.S. It's snowing heavily outside.  Come on weather, get a clue!  It's
> April!

Typical behavior for Boston.  Actually it was sunny and above freezing
again today.  The snow that had fallen around the crocuses has melted.

 --Steve
--
"And it tells the world Why should I wait,
This ice and snow gon' melt away." --Gillian Welch