[Mirror copy of document from ftp://naggum.no/pub/SGML94/;
poster delivered by Erik Naggum at SGML '94 at Tyson's Corner, VA]


How to Your Make SGML Documents Survive You

One excellent way to make sure your documents last at least as
long as they are valuable is to represent the information in
SGML, because as a language it enables data independence and all
that good stuff.  Like all languages worth knowing, however, it
allows a large number of forms of expression, and some of them
are well suited to immediate concerns, while restraint becomes
more important for longer-range goals.  This is a list of such
restraining conventions that you may want to consider for your
long-term information investment.

Inclusion Exceptions

    Don't use them.  If elements are independent of their
    position, collect them someplace else or let them be
    ordinary elements.  If elements are allowed "everywhere" in
    the document, make a parameter entity that allows them with
    PCDATA.  If they are allowed outside of text, explicitly
    include them in the content models.  Dynamic content models
    may be good for prototyping, but not for longevity.

Declared Content (CDATA and RCDATA)

    These "protect" & and < from markup recognition, but does so
    by making parsing much more complicated--it is not apparent
    from looking at an element that it is special.  Use entities
    or <![ CDATA [ marked sections ]]> instead.  The latter
    provide easily identifiable special treatment.

Potential Markup as Data Characters

    & and < are data when followed by digits or white space, and
    that's the only time they are safe.  Other characters may
    change their meaning when processed by a different system.
    Using </> as data when SHORTTAG is not used is a bad idea.

Non-obfuscatory Minimization

    * if the start-tag is omitted, omit the end-tag, too.
    * if the start-tag is "net-enabling", end with the net (/).
    * don't use net-enabling start-tags on empty elements.
    * don't mix empty start- or end-tags with omitted tags.
    * except with short references, don't let elements cross
      entity boundaries.
    * don't omit the refc (;) in entity references.
    * don't omit the tagc (>) in sequences of tags.

keep lines meaningful

    * don't use line breaks before the tagc (>).
    * don't use the line-end-suppressing entity reference.
    * don't use an entity reference for "new line".
    * don't use &#RS; and &#RE; in content.