Comment syntax in HTML and SGML

The syntax for comment declarations in SGML is

comment declaration =
	MDO ("<!"), (comment, ( s | comment )* )?, MDC (">")
comment =
	COM ("--"), SGML character*, COM ("--")

where the only delimiter recognized in a comment is COM ("--").

Paraphrasing, a comment declaration is

• MDO ("<!"),
• zero or more instances of: (
  • COM ("--"),
  • any text not containing COM ("--"),
  • COM ("--"),
  • zero or more whitespace characters ),
• MDC (">").

However, since most HTML user agents don't do the right thing, the safest syntax to use is:

• MDO, COM, ("<!--")
• any text not containing "--", "<", or ">"
• COM, MDC ("-->").

This generates comments that are legal SGML and will also be correctly processed by all but the most broken browsers.

If you're wondering why the comment syntax is so $!%#* screwy, it's probably because SGML comments were designed to be used in markup declarations (e.g., inside <!ELEMENT> and <!ATTLIST> declarations in a DTD). There the syntax makes perfect sense: comments both start and end with "--". Comment declarations (which are what you use in the document instance) seem to have been added as an afterthought.