Re: Single Sourcing and Hackos (long - sorry)

[Chris Despopoulos <cud -at- ARRAKIS -dot- ES>]

Thr problem with doing the Digest thing is that I wrote this
up before seeing that other people answered this point.
However, I'll send this anyway, since I went to the trouble,
and also I think it amplifies some points...  It may further
some understanding of this Generalized Markup domain...

Mary Deaton said:


     Chris, I am under the impression that Documentum
     uses SGML, which means that
     you can create a document description template to
     control formatting.
     Depending on which DDT you push the data into, it
     is formatted differently.
     As sub-sets of SGML, HTML and XML can be used in
     somewhat the same ways,
     although they are optimized for online delivery.

I don't want to seem too stern, but you are absolutely wrong
in your assessment of SGML.  A DTD has nothing to do with
formatting, and that is the point of it.  The DTD describes
the *structure* of the document information, which means it
describes the relationships between different units of data
in the document.  These relationships are hierarchical.

The original idea behind SGML was to make a format that was
absolutely independent and not proprietary.  For
independence from specific formatting programs, it had to
*not* represent formatting, because nobody can guarantee
that two different WP systems are able to implement the same
formatting.  Witness the Word to Frame conversion nightmares
for a graphic example.

It turns out that SGML is also a format disposed to
processing.  Because it contains meta-data, you can use it
to extract units from a document and do things with them.
Or you can have inclusions of external documents and data.
Or you can specify database queries from *within* the
document...  the list goes on.  It also turns out that
because the formatting is specifically NOT part of the SGML,
you can take the same information and format it in
completely different ways.  This is a first step to real
single-source material...  For example, in print, footnotes
show up at the bottom of the page.  For online there is no
page, so the footnotes appear inline.

But you can do more.  If you are up to it, you can create an
online formatter that imparts "behavior" to specific units
of data.  To carry on the example, now the footnotes are
links that pop up a little window when you click on them.
Or now the TOC is expandable.  Or now you simply exclude
data of a certain type because you know it is not viable
online.  Or now you simply show one path through a decision
tree, based on the navigation choices of the user.

Anyway, the "formatting" happens at the last possible
moment, and is handled by the application that displays the
SGML.  In the case of Maker+SGML, the formatting is applied
when you import SGML into the product.  Maker requires an
EDD that maps Maker formatting objects to elements.  Other
systems do this in other ways, but you generally need to map
the DTD to some sort of formatting representation of
structure.  (I have purposely omitted FOSSI.)

The point of an SGML document is that it includes a rigorous
description of what is "valid" structure in the document.
This corresponds (roughly) to the format of tables in a
database (if  I understand database terminology).  The
definition is accomplished via the DTD.  What it means is
this:  Given a data unit of a specific type (a specific
"tag"), you can know what other data must be included in it,
what other data may be its parent, and you can assign some
human meaning or value to the type of data...  For example,
a <SECTION userlevel="expert"> tag tells you something about
the data it contains.  The possibility that this tag can
exist in the context of the given SGML document is
determined by the DTD.

Finally, I would say that HTML cannot be used in exactly the
same way - for a number of reasons.  Obvious ones:  HTML is
primarily about formatting; HTML is one specific DTD, and
does not necessarily support the structural divisions that
actually exist in your text.  The latter point is like
saying HTML doesn't have enough tags, but it's more than
that.  One way to illustrate this limitation is to try
saving a Maker doc as HTML, then import the HTML back into
Maker.  On the way out, you have a many-to-one problem where
you specify many different formats that become one HTML
tag.  On the way in, you have a one-to-many problem, where
that one HTML tag *could* be one of many different
formats...  How do you know which one?  (This is why Maker
doesn't even try to give you an HTML import filter...  In
their wisdom they decided that would be useless.)  SGML is
sufficiently rigorous that you can indeed solve that
problem.  HTML, as one implementation (and religously
speaking, a heretical one) of SGML is not sufficiently
rigorous.

As for XML, I'm behind the curve on it, so I can't say
much.  However, I do believe XML is not meant to be
optimized for online, but rather optimized for ease of use.
However, I understand you can include a DTD with XML if you
wish, and it may be that to get the fullest processing
power, you should do so.

cud


 From ??? -at- ??? Sun Jan 00 00:00:00 0000=