The explanation is a bit unclear here -- the following more
detailed discussion should help clear things up.
<myElement /> and
<myElement></myElement> both correspond to an
'empty' element, so that these representations can in principle be used
interchangeably whenever an element instance has no content. However, 'for
interoperability' (with older SGML processors), the 'empty element' form
should _only_ be used if there is a document type declaration, and if this
formally states that myElement is empty:
<!ELEMENT myElement EMPTY >
This is because most SGML-only tools won't understand the empty-element
notation, and will look for an end tag for every start tag unless the
DTD explicitly denotes the element as empty (most SGML processors simply
ignore the trailing / in the empty element tags).
"Pure" XML tools are far more forgiving, so that any element instance that
happens to be empty can be written as an empty-element tag. Thus, the
following is entirely well-formed XML:
<myElement src="http://www.foo.org" />
<myElement> here is an element with content </myElement>
This may or may not be valid, of course, depending on constraints implied
by the DTD (if there is one).
As a convention, it seems reasonable to use 'empty-element' tags to
denote elements that must always be empty, since this notation makes it
less likely that someone would 'accidentally' put content in an EMPTY
element. But this is only a convention, and it has both advantages and
disadvantages - the disadvantages lying in the handling of this data
by older software.
Also, it is important to note that the
specification "canonicalizes" empty elements by converting them to
a start-tag/end-tag pair.