SGML was designed to be readable by computers, but it was also designed to be written by humans[Note 4]. It therefore has several features - collectively referred to as markup minimisation - designed to cut down the amount of markup you have to write.
The first such minimisation feature is tag omission: in certain
circumstances you may omit tags which are formally redundant. This
can happen only when the SGML parser can reliably infer the presence
of the omitted tags, and when the DTD author has permitted this. For
example, the content of the sect
element is a subhead element, followed by zero or more paragraphs (or
figures or tables), followed by zero or more subsect elements. That
is, the sect element contains the section, rather than containing
just the title (as is the case for HTML's H1 element, for
example). The subhead element
contains a title element, which
finally contains the text of the section heading. In other words, the
structure of a section is
<sect>
<subhead>
<title>Section 1</title>
</subhead>
<p>Paragraph text</p>
</sect>
<sect>
... <!-- and so on -->This would be tedious to type. However, the parser can infer the end of the section and paragraph elements (from the start of the next section and the end of the current section respectively), so those closing tags can be omitted; and since the section must start with a subhead, which must start with a title, those tags are redundant, too. That means that this can be compressed to just
<sect>Section 1 <p>Paragraph text <sect> ...
Many closing tags can be omitted, and a few opening tags, too (apart from subhead, about the only other one is px since it is the only permissable content of elements like abstract). If you include tags redundantly it doesn't matter (and may even make the text clearer in some cases), and if you omit them erroneously, the SGML parser will quickly tell you.
In cases where tags cannot be omitted, you can still cut down the
amount you have to type. Instead of typing <em>emphasised
text</em>, you can type just <em/emphasised
text>.[Note 5]
In certain cases, attributes may have only one of a limited set of
values. In this case, specifying one of these values is enough to
indicate which attribute you are referring to. For example, the
figure element has a `float' attribute, which
can take either of the values `float' or `nofloat'. You might
normally specify it as <figure float="nofloat">, but
since the attribute values come from a limited set, this can be
abbreviated to <figure nofloat>. The same is true of
the `export' attribute associated with exported IDs.
SGML also has single-character entity references. The character
~ in normal text turns into the entity
, and -- (two hyphens) turns
into &endash;. Such `short reference' characters are used
a little more in the programcode DTD (see Section 5).