Comments on MIME/SGML

["Daniel W. Connolly" <[e-mail]>]

                                                          Comments on MIME/SGML
               PROPOSED CHANGES FOR MIME REPRESENTATION OF SGML
                                       
                                          Daniel W. Connolly <[e-mail]>
                                                                               
SGML Entites in MIME

   I believe that the goals of Mr. Levinson's MIME Content-types for SGML
   Documents[1] are essential to the success of the intenet as an Integrated
   Open Hypermedia (c.f. HyTime[2]) system.
   
   However, after a careful reading of the SGML standard[3] (specifically
   section 6: Entity Structure), I believe that SGML/MIME[4] fails to specify
   the most important machine representation of an SGML document; that is, the
   SGML document entity (production 2, section 6.2).
   
   In conventional practice, an SGML document "doc" of type "T" is represented
   as a file doc.sgml, which looks like:
   

<!DOCTYPE T SYSTEM "t.dtd" [
<!ENTITY fig1 SYSTEM "foo.ps" postscript>
]>
<T>blah blah blah <figure graphic=fig1> blah blah blah</T>

   along with an SGML declaration in the file T.decl. Technically speaking, the
   SGML document entity is the concatenation of T.decl and doc.sgml (with
   perhaps some system-specific newline->RS/RE conversions). But according to
   the standard it's OK to interchange SGML documents with an implied SGML
   delcaration, and in practice, the SGML declaration is often compiled into
   the processing software.
   
   So for all intents and purposes, the file doc.html is an SGML document
   entity. And it seems critical that the file doc.html should correspond to
   the body of some MIME body part.
   
   The draft[5] misuses the term "DTD", aka "Document Type Definition" (defn
   4.104). An SGML document indeed has three parts: the SGML declaration, the
   prologue, and the instance. And the distinction between the term prologue
   and the term DTD is not trivial.
   
   First, to be pedantic, a DTD is not generally representable in SGML syntax.
   The concept of the DTD includes not only the SGML-representable formal part,
   but also the associated application conventions which cannot be represented
   in SGML. Also -- a document may have more than one DTD in its prologue.
   
   Second, to be practical, the conventional machine representation of a DTD is
   just an SGML text entity in the file t.dtd which looks like:
   

<!NOTATION postscript PUBLIC "-/Adobe/Postscript">
<!ELEMENT T - - (#PCDATA) >
...

   Note that it does not correspond exactly to the prologue of the document, in
   that it does not contain the <!DOCTYPE [ ... ]> markup.
   
   For these reasons, I suggest the following modifications to the proposed
   MIME representation of SGML documents:
   
   1. We make the following correspondence between the terms of the SGML
   standard and the MIME RFC:
   
      SGML notation => MIME content-type
      
       SGML SYSTEM identifier => MIME Content-ID
      
       SGML data entity (= notation, data)
      
       => MIME body part (=content type, body)
      
       SGML text entity (= sequence of characters)
      
       => body of MIME text/sgml body part (= seq of chars)
      
       SGML document => a MIME multipart/SGML body part
      
   2. We use text/sgml in stead of application/sgml for the SGML "files", since
   they are in general readable on teletype devices.e
   
   3. We change the parameters of the multipart/SGML content type from
   

               sgml-part       := "intance" / "declaration"
                                / "dtd" / "fosi" / extension-token

   to
   

               sgml-part       := "document" / "declaration"
                                / "dtd" / "fosi" / extension-token

   where "document" is required, declaration is optional, and dtd is acutally
   redundant (since it's in the document entity) but useful, since a MIME UA
   might want to know what kind of document it is without parsing the document.
   
References

        Network Working Group, Internet Draft: MIME/SGML,
      <draft-levinson-sgml-01.txt>, E. Levinson, Accurate Information Systems,
      Inc., January 17, 1993
      
        ISO 8879:1986, Information Processing: Text and Office Systems:
      Standard Generalized Markup Language (SGML)
      
      ISO/IEC 10744 Information technology -- Hypermedia/Time-based Structuring
      Language (HyTime)
      
Editorial Note

   I used HTML because it works to a certain extent, not because I think it's
   exactly how I think internet IOH should work. My comments on HTML are still
   under development. See my notebook on the design on an HTML successor[6].
   
Production Note

   This document is brought to you by the following tools:
   
       Lucid emacs
      
       html-mode by Marc Andressen
      
       SGMLs
      
       NCSA Mosaic