==================================================
Proposal for a language optimized for WWW delivery
==================================================
This proposal suggests a strategy for evolving HTML by using two
different forms of markup, one for authoring and one for delivery.
The first format, optimized for simple, platform-independent
structural markup, would be an incremental improvement of HTML 2.0
that would add a few basic features such as tables and turn off
features that are deprecated in the current 2.0 specification. This
authoring language, HTML 3.0, would continue to fulfill the goal
described in the IETF HTML specification: "a simple markup system used
to create hypertext documents which are portable from one platform to
another."
The corresponding delivery language would be based on an existing
online delivery format called SDL that was developed for the Common
Desktop Environment (CDE) help system. Extended for WWW delivery use,
the proposed language would be called HDL (Hypertext Delivery
Language).
--------------------
Client/server markup
--------------------
Networks today are based on the fundamental realization that certain
functions are appropriate to servers, while other functions are
appropriate to clients. The proposal set forth here is based on a
similar principle: that certain forms of markup are appropriate to
authoring and other forms are appropriate to delivery.
All systems of markup represent a compromise between expressive power
and ease of authoring. As the number of markup elements (tags, etc.)
in a language increases, its ability to convey structural and
presentational information increases in proportion, but so does the
difficulty of creating documents in that language. The break-even
point between these two opposing tendencies depends on the
availability of suitably powerful authoring tools. For example,
domain-specific industrial markup languages such as DocBook (software
documentation) and ATA-100 (aircraft documentation) are typically very
rich in their ability to convey structural information, but are almost
unusable without the help of expensive structured authoring systems.
HTML, on the other hand, is a markup that is close to the upper limit
of what can reasonably be produced using minimal authoring tools such
as simple text editors and understood by relative novices.
Currently, HTML is both an authoring language and a delivery language.
As use of the Web matures and control over online delivery becomes
increasingly necessary, however, HTML's limitations as a delivery
vehicle will become more apparent. We are already seeing the
implementation of uncontrolled, nonstandard additions to the HTML tag
set in order to overcome these limitations. Such ad hoc changes to
the standard not only raise the spectre of chaos as commercial Web
delivery systems compete for market share; they also threaten the
usability of the HTML language itself as a straightforward medium for
the expression of ideas.
The best way to avoid multiple incompatible delivery languages and the
loss of interoperability is to standardize on one system of markup
capable of expressing the complete range of typographical effects
needed for cross-platform online display and to convert to this
delivery format from other languages optimized for authoring or
domain-specific retrieval. This proposal leverages an existing
standard, SDL, that is already available to serve as the basis for a
common Web delivery language.
------------------------------------
The Semantic Delivery Language (SDL)
------------------------------------
SDL was designed by Hewlett-Packard as a language for delivering
online information, in particular help pages, to users of X-based Unix
desktop systems. It is the primary delivery medium for the COSE
Common Desktop Environment (CDE). The SDL standard is maintained by
HP's Workstation Systems Division and is jointly owned by HP, IBM,
Sun, and Novell. Digital, Hitachi, and Fujitsu will soon join this
list. The standard can be obtained by anonymous FTP from
ftp.ora.com:/pub/davenport/SDL.
SDL has the following characteristics.
* Uses a minimum number of elements (54 in all)
* Suitable for many forms of online information
* Designed for distribution and display (not authoring)
* Parses quickly
* Displays quickly
* Allows fast access to individual pieces of documentation
* Supports infinite structural depth
* Supports multibyte text
* Supports advanced retrieval functions through the assignment of
search weights and the retention of original semantic and structural
information
* Enables viewers to run on systems with relatively small amounts of
memory
In addition, SDL provides a rich yet efficient set of structures for
controlling nearly every aspect of display that is relevant to
device-independent online delivery in cross-platform environments,
including:
* Wide variety of fonts
* Special characters
* Lists and tables
* Graphics
* Flowing text around graphics
* Captions on graphics
The typography of any block of text can be directly controlled in SDL.
Typographical controls built into the styles component of the language
include font size, font width, font color, font style, font spacing
(monospaced or proportional), font weight, font slant, left margin,
right margin, top margin, bottom margin, border type, vertical
justification, and horizontal justification. In addition, heading
elements can be placed in any one of eighteen different positions with
regard to the block of text with which they are associated and
likewise for the positioning of subheads in relation to their
associated heads. Non-typographical attributes include TOC level and
search weight. All of these attributes are optional and will be
supplied with defaults appropriate to the operating environment if
omitted. Specified attributes are taken as "hints" that will degrade
in a controlled way depending on the level of support for the
necessary functions in the operating environment.
Like HTML 2.0 and most other standard markup languages, SDL is an
application of ISO 8879 (SGML).
----------
SDL to HDL
----------
SDL lacks several features that would need to be added in order to
create a Hypertext Delivery Language for the Web. These include:
* Forms (in the HTML sense) for data input from a remote user
* A horizontal rule element (TBD)
* Embedded widgets
* Explicit style attribute switching based on the operating
environment so that different font names (for example) could be
specified for different operating systems
* Extension of the existing conditional text mechanism to non-Unix
systems
It is to be expected that an HDL standardization effort would discover
the need for further extensions to SDL beyond those listed above.
---------------------------------------------
HDL support outside the current WWW community
---------------------------------------------
Sun has announced that it will be supporting SDL in a Web-capable
browser that will ship with its Unix operating systems. Since the CDE
1.0 code is jointly owned by the other COSE members, there is a very
high probability that some or all of them will also ship SDL Web
browsers. A proposal for a universal SGML browser that will support
SDL directly (without intermediate translation of SDL style
information to another stylesheet language) is currently under
consideration by an important industry consortium and seems likely to
be approved. SDL is a natural choice for help systems and will likely
achieve widespread use as a delivery mechanism for Unix help and in
some quarters as a replacement for Winhelp. All of this is
independent of whatever support might come from the current WWW
community as a result of this proposal.
------------------------------
Availability of reference code
------------------------------
HP has generously offered to make the source code for the X Windows
SDL viewer used in CDE 1.0 publicly available if a suitable
educational institution will take responsibility for distributing and
supporting it. The SDL browser will still need the addition of the
features noted above in order to become the cross-platform HDL browser
envisaged by this proposal, but obviously the public availability of
the SDL parsing and rendering code will make the implementation of HDL
viewers, and the incorporation of SDL capabilities in existing HTML
browsers, much easier than it would be otherwise.
------------
Action items
------------
It is hereby proposed that the IETF HTML working group adopt SDL as
the basis for a future Web browser language (HDL) and work to define
the additional features that would make HDL a workable standardized
alternative (not a replacement) for HTML in future WWW browsers.
It is further proposed that work on HTML 3.0 continue in order to
further refine a format-independent markup language suitable for hand
coding documents intended for Web delivery.
Since many document producers who could benefit from the adoption of
an HDL standard must know for planning purposes whether such a
standard will become part of HTML browsers or whether they should
begin working toward solutions based on universal SGML browsers
instead, the HTML working group is urged to give this proposal its
earliest attention.
-------------------
Further information
-------------------
An accompanying document contains a further discussion of the reasons
for supporting this proposal.
========================================================================
Jon Bosak, Novell Corporate Publishing Services jb@novell.com
2180 Fortune Drive, San Jose, CA 95131 Fax: 408 577 5020
A sponsor of the Davenport Group (ftp://ftp.ora.com/pub/davenport/)
========================================================================