The HTML Professional DTD

Background

Derivation

The first HTML Document Type Description was devised to support the original versions of World-Wide Web software in the early 1990s. A revised and more widely-discussed version was codified as HTML 2.0 by the Internet Engineering Task Force Working Group on HTML, and adopted by the IETF as a draft standard, RFC1866, in November 1995.

A more advanced but experimental version, HTML+ (now obsolete), had been under discussion for several years, and many of its features were republished as HTML3 in an Internet Draft (March 1995). On expiry, this draft was taken back into discussion, and passed to the World-Wide Web Consortium (W3C), when it took over HTML development from the IETF in 1996.

The W3C has attempted to reconcile the sometimes conflicting aspirations of some of its members by publishing two experimental versions of HTML in May and July 1996: respectively Wilbur, the codename for HTML 3.2, which is largely HTML 2.0 disguised with the addition of stylesheet and scripting support; and Cougar, which is a less structured version of HTML3 with some changes to the tables model, and some of the new content markup removed. While both have their adherents and proponents, both are highly selective, and neither can be said to be in any way stable or comprehensive. Both also suffer from a lack of discipline, because they try to accommodate historical, even obsolete, practice in a fully backwards-compatible manner. The result is that both use a flat content model for the body of the document, and have reintroduced --- without warning or explanation --- several elements which the original designers had clearly marked for removal.

Companies implementing Web software (particularly browsers), and individuals implementing HTML pages, have throughout this period continually sought additional markup facilities for their own and their customers' purposes. In some cases element names were simply invented on an ad hoc basis, with no attempt to define when or how they can be used; in other cases, market position was available as leverage to add support for them ahead of their inclusion in a DTD. While the former method can be excused in individuals on grounds of ignorance, the same argument cannot be made for corporations. The latter method is natural from a marketing point of view, but it has led to a further Balkanization of the DTD position, with separate versions needed to describe differing implementations by Microsoft, Sun, Netscape and others, as well the implementation of more practical versions by SoftQuad and other editor makers.

HTML Pro

The edition of HTML presented here, codenamed Aardvark, is a composite of all known versions to date. It contains all the elements published (as far as they can be found or identified) in the various forms of HTML since 1990, in a manner which can be used by editors, browsers, parsers, databases, search engines, formatters, indeed any conforming application of ISO 8879 --- Standard Generalized Markup Language (SGML, the language in which HTML is written).

HTML Pro is intended for use by the information professional who has neither the time nor the inclination to participate in the interesting but destructive competition between rival browser makers, and who is unwilling to mislead her clients by creating anything other than valid, conformant instances of HTML.

The DTD

There are now 140 elements in the DTD sharing 245 different attribute names. Of these attributes, seven are common to all elements (see below). The element with the greatest number of attributes is INPUT (52).

Downloading and installing

Downloading

The HTML Pro DTD is available in several forms: this file with active links is maintained at http://www.arbornet.org/~silmaril/dtds/html/htmlpro.html and the distribution files are held at ftp://ftp.ucc.ie/pub/html/

• The full distribution as a single compressed archive for UNIX (htmlpro.tar.gz), PCs (htmlpro.exe), or Macs (htmlpro.zip.hqx), and a VAX/VMS save-set version is in preparation (htmlpro.sav);

• A minimal installation (panofkit.exe) consisting of the DTD, stylesheet, navigator, and support files to demonstrate the use of a free SGML browser/viewer for PCs (Panorama: installation instructions below).

• The htmlpro.dtd file on its own, for existing users who are upgrading, or who don't want or need the background and supporting files;

• This document (htmlpro.html), also available for use with Panorama (htmlpro.sgml, see below), and as a PostScript file (htmlpro.ps) for printing or viewing;

The full or the minimal distribution should be saved to a suitable temporary directory before installing (unless you are running WinZip, which automates this). There is a DOS htmlpro.zip version available for non-Windows users.

Installation

The UNIX version unpacks with the commands:

gunzip htmlpro.tar.gz
tar -xvf htmlpro.tar

UNIX users who intend to use Emacs with psgml-mode to parse SGML Public Identifers in their files will need to create downcased soft links to the subdirectories DTD and ENTITIES within any of the Public Owner directories in /usr/local/lib/sgml, for example with the commands (in each directory needed):

ln -s DTD dtd
ln -s ENTITIES entities

The PC/Windows version is a WinZip self-installer which unpacks automatically when you double-click it. If you prefer not to execute such programs, you should download the DOS version instead and unzip it manually.

Emacs users under DOS and Windows do not need to change any directory names.

The Mac version is a BinHex'd zip file which can be unpacked by UnStuffIt or similar dearchiving program. The BinHexing should be unnecessary, in which case you can download the DOS zip file referred to above, and drop it onto UnStuffIt. Macs are poorly catered for by the SGML software community, so little information is available about where files like DTDs and character entity collections should be installed: if you have this information, please mail us.

Versioning

The versions of HTML included in this edition were taken from public copies of DTDs and fragments on the Web: this list appears in slightly altered form later on, showing the codes used in the DTD to identify sources.

My thanks to all the many authors and contributors to these DTDs, whose notes and comments have made it easier to work out what to do, and whose research has made it possible to identify to some extent which browser can support which elements.

The current version of the HTML Pro DTD is 0.11, and it is released for comment by the relevant interested parties on the www-html mailing list and elsewhere. Please forward all comments and suggestions to the mailing list for discussion.

Basic changes introduced

Very few changes to structure have been made, although those familiar with the internals of previous versions will notice the large amount of additional material from the other sources, including the unilaterally-introduced elements invented by some browser makers (which in many cases has involved a guess at their structure, as the companies have often been reluctant to provide details).

The header is as defined in HTML3, with the addition of NOSCRIPT and BGSOUND. The content model has been rearranged to allow intermingled META and LINK elements, and still try to help editing systems keep header elements in some semblance of logical order (my thanks to Joe English for this code). The use of META is now expanding to include the Dublin Core metadata qualifiers (Title, Subject, Author, Publisher, OtherAgent, Date, ObjectType, Form, Identifier, Relation, Source, Language, and Coverage).

The most significant structural change in the text body is to the way in which content models are used. The original mechanism was for %body.content; to contain both structural and descriptive elements as peers, so that there was no distinction between, say, a list and some running text in italics. The implication was originally that inline markup, which was identified as %flow; in earlier versions, should be contained in a paragraph-level element, and this is in fact what SoftQuad's HoTMetaL Pro does when it imports existing invalid HTML.

The DTD has not been modularised in any way: on the contrary, it is currently an entirely flat file with only the barest minimum of entities. This is deliberate at this stage, as some non-SGML tools were used in the conversion of the code from the various sources, and following the text formatting established by Near&Far for the layout proved useful for this purpose, where each declaration and each attribute definition is followed by a line containing a comment. In a future version, use will be made of the work of Altheim, Connolly, Maler, and Allen in defining the modularised version of HTML.

There are now four classes of elements, defined by the entities following:

The distinction is that the insertions class can be peer with text as well as with structure, whereas text can nest only within structure.

Some changes have been made to the content models of the text elements to accommodate this, but the only noticeable ones are SUP and SUB, which may now contain math elements even when used in non-math mode. The exclusion exceptions for MATH have been significantly reduced because it would appear from discussions with mathematicians that it is in fact unobjectionable for math to contain many of the elements previously proscribed. The effect of these changes is an increase in the amount of mixed content (see note below), which some analysts regard as pernicious.

The exclusion exceptions for PRE now include TT and BR, as neither have any relevance in preformatted fixed-width material, but SUB and SUP are now permitted, as it seems perfectly reasonable that an author might want to represent typewritten material containing subscripts and superscripts. PRE can now include IMG to enable character-cell-sized image glyphs, and APPLET, EMBED, and OBJECT to allow the dynamic generation of fixed-width content.

Accessibility issues

The ICADD fixed attributes which were added by the late Yuri Rubinsky have been reinserted for those elements to which they were attached in RFC1866. Work is proceeding with the developers of the ICADD DTD to identify what equivalents to add for the remaining elements, and this is being coordinated with a similar exercise being undertaken on the Text Encoding Initiative (TEI) DTD. A note to RFC2070 says:

Following the informal discussions at SGML'96, some recent suggestions from Harvey Bingham have been incorporated in respect of Tables (additional attributes).

Recent alterations and additions

I am grateful to Foteos Macrides for keeping me updated on the implementations used for Lynx and other browsers. His request for a TITLE attribute for FORM and MAP triggered a rework of the common attributes which are now applied to all elements:

Internationalization

The LANG and DIR (and some changes to the SGML Declaration) are reintroduced directly from the Internet Draft on Internationalization of HTML (now RFC2070). Authors should note the following (from RFC2070) and ensure that they use only valid numeric character references:

The relevant RFCs referred to here and in the DTD source code are included in the distribution in /usr/local/lib/sgml/IETF/HTML for want of a better location.

Indexing

Scott Preece suggested the NOINDEX attribute to provide a means of allowing an author to specify to indexing engines that the document is not considered meaningful for index-gathering (valuable for pages originally generated from scripts but stored as temporarily static files).

HTML3 revisited

Foteos also brought to my attention the need to replace more of the material originally in HTML3 which had fallen away, as well as the attributes supported by Lynx, including the DISABLED attribute for INPUT, OPTION, SELECT, and TEXTAREA; the PLAIN attribute for UL; CONTINUE for OL; ALT for BGSOUND, APPLET, AREA, IMG, and EMBED; and the HREF attribute (and ACTION as a synonym, from Mosaic) for ISINDEX.

<isindex http:foo.bar.com/blort>

Drazen Kacar suggested the reinclusion of the CLEAR attribute on all structural elements, so this has been done. The assumption is that it is not needed on inline or descriptive markup.

Embedded objects

The controversy over EMBED vs OBJECT was no less acrimonious than the earlier one over APP or APPLET vs FIG, and has been treated by including all five elements as implemented. The now-expired Internet Draft on Compound Documents by Paul Burchard and Dave Raggett included PARAMS as an attribute on EMBED, rather than the attribute soup approach piloted by some browser-makers, and the W3C now has their own document on the subject.

Mathematics

Little comment has been received to date about math content. The view was expressed at SGML'96 that the whole question of SGML math was still up for grabs, so it is unlikely that any significant change to the HTML3--based content will be made unless mathematicians present some kind of conclusion.

I am grateful to Dave Carter, however, for the suggestion to include FONT in the model, so that manual/visual adjustment can be made to large symbols where browsers fail to render them automatically. There is as yet no default inclusion of the ISOams* character entities (although that can easily be done with a simple edit).

There is one significant new attribute, ROLE on MATH, which can take the values INLINE or DISPLAY. This corresponds with standard mathematical usage as reflected in TeX (this attribute has been made #REQUIRED). Following hints from Tom Magliery, it is possibly true that most math authors are writing in TeX and making GIFs.

TV browsing

There has been one special group of added elements: those implemented by WebTV without a DTD. The elements added are AUDIOSCOPE and SIDEBAR in the insertions class, and BLACKFACE, LIMITTEXT, NOSMARTQUOTES, and SHADOW to the text class. A substantial number of specialist attributes have been added to handle the WebTV output, and these are identified by (T) in the DTD source.

Plugin support

The only requested addition has come from FirstFloor Software, who provide the Smart Bookmarks module shipped in Netscape's PowerPack. This allows for specification on any A (anchor) element of the date and time (HTTP format, eg Sun, 06 Nov 1994 08:49:37 GMT) of a bulletin, plus displayable text and links to the bulletin and an image.

Manufacturers who are experimenting with additional markup are invited to send details (with a DTD fragment) so that the elements can be included in a subsequent release of HTML Pro. The address for submissions is silmaril@m-net.arbornet.org

SGML experiment

The astute reader will have noticed the new elements ELEMENT and ATTRIB, which have been added to this version to test their use in documentation, in order to accompany the already proposed ENTITY and COMMENT. It is not the intention that they should remain past v1r0 unless a formal approach is made to the then controllers of the HTML standard.

The COMMENT element is in any case wholly redundant, and it is assumed that the element was invented in an uncontrolled burst of misdirected enthusiasm, coupled with a significant failure to read the HTML specification, as COMMENT does not correctly implement comments at all (embedded markup gets interpreted instead of ignored, in implementations to date).

Documentation of markup sources

Where it has been possible to find out who invented an element, this is noted in the comment which follows every element and attribute in the DTD source code. This comment is in the form used by Microstar's Near&Far which they call a title:

<!element foo - - %bar;
 --<Title>(X)comment goes here-->
...
<!attlist foo %commonatts;
 bar CDATA #IMPLIED
 --<Title>(Y)comment goes here-->

The attribution is given by quoting one or more single-letter codes in parentheses immediately before the comment (see Figure 2 for the codes). I am grateful to the authors of the WebTV proposals for their work on identifying some of these origins.

The HTML3 concept of HTML-specific character entity files has been ditched, and this version includes the whole of ISOlat1, ISOlat2, ISOnum, ISOpub and ISOtech, which should have been done years ago.

Mixed Content

The infamous problem of mixed content in list items (and elsewhere) has been tackled head on by simply permitting it: text data is allowed as well as paragraphs and other structural markup like lists, tables, and forms. This may cause some less well-endowed editors a little grief, but the advantages of being able to tweak the performance characteristics of various browsers are too great to pass up.

This has been regarded as unavoidable, given that different usages are needed, and the same has been permitted within the cells of tables, as established practice dictates. Rather than produce several DTD versions, one permitting text where another permits only further markup (the lax vs strict versions of many HTML DTDs to date), it has been assumed that professional users would be equipped with a suitable editor capable of handling the behavioral and formatting requirements of mixed content.

Usage

It is assumed that the user is aware of the meaning and usage of the elements of HTML, and of the principles of SGML. Those who are unfamiliar with these will find help in the only SGML-compliant book on the use of HTML, The World-Wide Web Handbook by Peter Flynn (ITCP 1995, ISBN 1-85032-205-8), which covers HTML 2.0 and HTML3. The Reference Card which accompanied the book is also available on the Web and for download as a PostScript file to print.

A more extensive list including the proposed additional elements, with a short comment on their intended usage, is included in the HTML Pro distribution as file htmlpro.tag (as used in SoftQuad's RulesBuilder), and is reproduced here in Figure 3.

Outstanding items

The machine-generated status of the DTD file meant that there was a substantial amount of legacy comment from the assorted versions used in compositing the DTD. The majority of this has now been edited out so that obsolete comments and parameter entities are not left to confuse the unwary reader, but users who locate undetected residua are asked to report them (to the address below).

Now that the DTD is stabilizing, there is a need to re-parameterize some of the material, and use some of the architectural material already tested in the modular version of the standard.

The elements for which no ICADD fixed attribute exists need analyzing and the relevant values adding from the International Committee for Accessible Document Design DTD ("-//EC-USA-CDA/ICADD//DTD ICADD22//EN"; work is under way on this).

A longer-term goal is to devise and maintain publicly a table of which elements and which attributes are actively supported in which browsers.

There are undoubtedly other errors, both of omission and of commission, and I would be very grateful for details so that they can be fixed: silmaril@m-net.arbornet.org.

Distribution

The current distribution includes the following files. The default installation directory is given in [square brackets].

• Directories containing known source DTDs in their Public Text versions. The directory names and the files they contain are those expected by an SGML-compliant system expanding the Formal Public Identifier (FPI) in the manner of sgmls. The directories are [/usr/local/lib/sgml/...]:
  IETF/
  ISO_8879-1986/
  Microsoft/
  Netscape_Comm._Corp._Strict/
  Netscape_Comm._Corp/
  Silmaril/
  SQ/
  Sun_Microsystems_Corp/
  W3C/
  WebTechs/

  On a UNIX or DOS-based system, this usually means the mapping given in Figure 4. The somewhat strange names of some existing files are a result of a misunderstanding of the nature of the role of the ISO 9070 Public Owner Identifier by the authors of some of the DTDs.

• This document (htmlpro.html) in HTML format [/html/dtds];
• This document in PostScript format (htmlpro.ps) [/html/dtds];
• The HTML Pro DTD in a short-filename form (htmlpro.dtd) [/html/dtds];
• One-line explanations of the elements (htmlpro.tag) suitable for input to compilers such as SoftQuad's RulesBuilder [/html/dtds];
• Multipage table of elements against attributes showing which uses what (htmlpro.lis, a 66--line by 132--column lineprinter listing) [/html/dtds];
• Copies of the principal Internet Drafts and RFCs referred to above [/usr/local/lib/sgml/IETF/html/];
• Precompiled (binary) versions of the DTD for some application systems:
  SoftQuad Author/Editor rules file (htmlpro.rls) and sample style file (htmlpro.stl and htmlpro.asf);
  Emacs psgml-mode CED file (htmlpro.ced);
  Microstar (Near&Far) Binary Format file (htmlpro.mbf).
  [/html/dtds];
• Sample DSSSL skeleton structure (htmlpro.dsssl) generated with psgml-dsssl [/html/dtds];
• Sample style sheets and navigators for Panorama Pro (htmlpro.ssh and htmlpro.nav), plus catalog and entityrc file fragments [/Program Files/SoftQuad/Panorama PRO/...];
• Sample style sheets and navigators for Panorama Free (htmldemo.ssh and htmldemo.nav), plus catalog.fre and entityrc.fre file fragments [/Softquad/Panorama/...];
• Graphical diagrams of the DTD structure (htmlpro*.eps) taken from Near&Far [/html/dtds];
• Assorted icons for use with the documentation when displayed with a browser [/html/dtds].

Software manufacturers who would like compiled versions in their own format included are invited to contact the authors.

Silmaril//

HTML Pro v0r11 19970101//

SGML support

No SGML software is included, as this can all be obtained online. A good list of software and associated information on projects and activities is kept by Robin Cover at http://www.sil.org/sgml/sgml.html.

Editing

The minimum to get started editing HTML in a fully-conformant manner is an SGML editor. Some of the most popular cross-platform systems seem to be:

• Author/Editor from SoftQuad. This is a straightforward graphical editor for any DTD, and is available for MS-Windows (3.x and 95), Apple Mac, and UNIX/X. It is only an editor, not a DTP system, so its paper-publishing facilities are limited (SoftQuad do a separate system called Sculptor which takes A/E into the professional publishing field).

• GNU Emacs with psgml-mode. This free editor is well-known in the UNIX field (it also reads mail and Usenet news, browses the Web, does FTP, and a thousand other things as well as just edit), but it also works well under MS-DOS and Windows, and there are user-supported versions available separately for Macs and VMS.

Browsing

Softquad also do a Web browser add-on for Windows PCs called Panorama. There's a free version online, but if you're planning on using SGML more than trivially, the commercial version, Panorama Pro, is strongly recommended.

This file can be used with either version, but with PanoFree you need to install the minimal subset of HTML Pro, whereas PanoPro it will automatically download the right bits and pieces for you. If you've got PanoPro already installed and working, you can go right ahead and click here.

If you want to get started with the free version, do the following:

1. Make sure your regular browser (Netscape, Mosaic, Explorer) is already running first: presumably it is, or is at least possible, or you wouldn't be reading this (unless someone else printed it off for you);

2. Download PanoFree and double-click to install it, making sure you use the default directory (C:\Softquad\Panorama). If it doesn't mention your browser as one it works with, don't worry --- after installation is finished, go to your browser's Options menu and reconfigure your Helper applications for the MIME type text/x-sgml to make file types .sgm and .sgml start the c:\softquad\panorama\panorama.exe program;

3. Close down and restart your browser so that it picks up the link to Panorama correctly;

4. Don't be tempted to try and run Panorama yet before the installation of HTML Pro is complete;

5. Download the PanoFree HTML Pro starter kit (about 45Kb) and double-click to install it;

Now you should be able to click here and view this file using all the facilities of a real SGML application (well, nearly all: some of them are reserved for customers of the commercial version, Panorama Pro).

Disclaimer

Silmaril has no connection with or financial interest in SoftQuad, Inc. or with any other company mentioned in this document.