I think you're mistaken.
What evidence did you base this conclusion on? Here's some evidence
that shows that the current HTML 2.0 DTD permits this:
[Whether it should or not is open to debate... ]
connolly@ulua {** NONE **}../html-test[1037] cat test.html
<!DOCTYPE HTML PUBLIC "-//W3O//DTD WWW HTML 2.0//EN">
<title>Can INPUT go in PRE?</title>
<form action="xxx">
<pre>
Enter xxx here: <input>
</pre>
</form>
connolly@ulua {** NONE **}../html-test[1041] SGML_PATH="%N.dtd:%N.sgml" sgmls h
tml.decl test.html
(HTML
(HEAD
(TITLE
-Can INPUT go in PRE?
)TITLE
)HEAD
(BODY
AACTION CDATA xxx
AMETHOD IMPLIED
AENCTYPE IMPLIED
(FORM
AWIDTH IMPLIED
(PRE
-Enter xxx here:
ATYPE IMPLIED
ANAME IMPLIED
AVALUE IMPLIED
ASRC IMPLIED
ACHECKED IMPLIED
ASIZE IMPLIED
AMAXLENGTH IMPLIED
AALIGN IMPLIED
(INPUT
)INPUT
)PRE
)FORM
-\n
)BODY
)HTML
C
Dan