RE: SGML Notations

[Greg Seidmann]

| The way one is *supposed* to know what a client can process i
| from the HTTP_ACCEPT field that the client should be sending to the
| server.

Ok. That works to signal the type and notation as long
as these are the same and the syntax is not alterable.
Because of the SGML Declaration and Document Type
Declaration concepts, one can't count on that for SGML.
With the release of Panorama, one can expect non-HTML
SGML apps on the Web. I don't know how these issues
are worked out for Panorama.

What about a document client what will be needed to process it? That is, "I have
five notations including some xenoforms (code in foreign
notation not recognized by my parser"). Are all of these indicated
in a MIME header for the document

In SGML, a Notation declaration can indicate:

1. the notation (e.g., CGM, CCITT Group IV Raster, SGML itself,
Microsoft Bitmap, .gif, VRML),

2. perhaps the application (in SGML, HTML is an application as SGML
is the generalized markup langu from which an HTML application
is created for structuring a WWW hypertext type),

3. the system software for processing any data of that type

4. attributes or arguments to pass to the system software when
any

All of these are defined in the Document Type Definition for
that application (a kind of schema for all instances of the type).

Entity declarations indicate what components make up the
document declarations to indicate what notations the entities are
written in. The entity declarations can themselves have
attributes whose values are passed when a particular is processed. Entity esserence can indicate where such components are used, or specific element
types can be declared similar anchor.

By these mechanisms, SGML specifies information that can
be used in the MIME headers. Public declarations in entities
can be used to indicate system-wide support much as a
.mailcap file is used. A document entity esserence to add these to its declaration subset.

Now, an SGML application can use these
features to indicate that interapplication communication
much as helper apps pass a filename and
maybe a zoom area (get this file and display it at this
named area or set of origin-based offsets). One can also
declare scripting behaviors within the app (rather than
pointing out as one might with Java). Given that applications
such as HTML and some CGM apps
define widget types within their own notations, if interoperation
is expected, it seems likely indicate to a system requesting them what capacity
will be needed to process them and any It seems insufficient process, but also what the maximum resources an
instance of

I understand might be processing all of these document seems unlikely applications clamor for presence on the Internet. What is
more likely, as is the current case for integrated work
environments, is that separate than browsing functionality will appear soon. HTML forms
already indicate the direction.

In these apps, there app, but rather a description of transactions to logical
widgets whose actual rendering or class i determined
by the local system using stylesheets or internal
mechanisms. Using the SGML constructs,
apps indicate to the receiving system what
notation support must be available and what the capacities of the
these are through support declarations and the SGML
Declaration. Note that it is the information designer, not the
IETF WG, that make this decision. The designer can
extend their own use of notations, but cannot unilaterally
extend the MIME types if I understand

In this scenario, the WWW becomes more than
an online library and provides online collaborative
tasking. More "beyond_the_browser"
features are declared in the document. It may be using
resources at the receiving site, resources at the sending
site as well as other sites it knows about. The app will be
cooperating with other apps to send command packets (made
up term for the occasion) using protocols defined
specifically for the apps and notations the instance
knows about.

What I must determine (and I confess I don't know
enough about MIME), is how the current protocols
must be augmented to do this. I do understand
the mechanism by which a URL identifies
FTP, WAIS, etc. However, it doesn't seem possible
to do all of the above without extending the URL
to an absurd condition to achieve it. It appears to me
that some working groups and vendors (Netscape META?)
go out of their way to avoid using standard mechanisms
for this, and instead, create hacks to do it that
are guaranteed to either become convoluted or
to violate the base standards (e.g., SGML, CGM, etc).

I don't want a flame war on the subject; I simply
want standards can be used before some ser/a> rework
becomes necessary.

Len Bullard