Most likely a cookie bug - reassigning there.
I think I can can answer this. I don't know if it was a cookie bug or not, but there was a period when cookies were replicating themselves---each time the browser was run, the cookies were duplicated. The existance of multiple cookies confused the browser-buster page counter, and it generated a NaN cookie value. The bug was eventually fixed and duplicate cookies disappeared. But if you had picked up the NaN, it stayed around (NaN composed with anything is still NaN). As far as I can tell, the only way to get rid of it is to edit the cookies file and delete the line (while the browser isn't running, of course). If there's a residual bug, it's that a malformed cookie file with two values (on two lines) for the same cookie aren't cleaned out.
<!-- --> make an HTML comment. My guess is that the script is showing up in the <HEAD> of the document in one case, and in the <BODY> in the other, and this is exposing something different in Mozilla about how we bring document or document.forms into the scope. Reassigning to the DOM component.
The null error problem was introduced and fixed by pollmann earlier this morning. We're still getting a script error (document.forms has no properties) on the page count page http://komodo.mozilla.org/buster/count.html. The reason is that the document.write of the <FORM><INPUT...></FORM> results in the parser (actually the NavDTD) holding on to the tags in its misplaced content list (since a BODY hasn't been explicitly opened). Passing this one along to harishd - seems like we might need to flush the misplaced content list if the content comes from a document.write.
*** Bug 24479 has been marked as a duplicate of this bug. ***
Adding Clayton to the CC list.
FIXED by opening up BODY for INPUT ( emulating IE to an extent ). BTW, this fix is quirks mode only.
Verified with 2000-02-02-14.