As far as I can tell, it's the rowspans conflicting with the fact that when hidden there are no rows to span. Also you might want to use "display: table-row" rather than block.

And you don't seem to have any colspans despite having a mismatch in column numbers - is this not needed anymore?

So at a guess, you'd have to make the javascript add in/remove the rowspans as appropriate.

As for the javascript not making rows show/hide in ff... what's rclass? I've never heard of it and ff isn't picking it up (it's always undefined). If you switch rclass with id it works fine.

thisRow = rows[i];
if (thisRow.rclass == classId)
{
	x += '[i:'+i+']';
	if (thisRow.style.display == 'none')
	{
		thisRow.style.display = 'block';
	}
	else
	{
		thisRow.style.display = 'none';
	}
}
else
if (thisRow.id == classId)
{
	for (n in thisRow.childNodes)
	{
		if (n != 'length')
		{
			thisCol = thisRow.childNodes[n];
 
			if (thisCol.className == 'td_all')
			{
				if (thisCol.crowspan == thisCol.rowspan)
				{thisCol.rowspan = 1;}
				else
				{thisCol.rowspan = thisCol.crowspan;}	
			}
		}
	}	
}

Not sure what you're trying to do with that code - does rclass now work. If so, why check for both rclass and id? If not, crowspan won't either, will it?

And wouldn't it be better to work out the appropriate rowspan in js?

I feel like I'm missing something so I probably am...

That code isn't triggering for anything because ff won't pick up 'rclass' (I assume it would if you were working in pure xml) and there are no appropriate ids set..?

(hug)

If you really want to pass the target-rowspan through a tag, maybe you could pass it through something valid where it won't do anything, like 'align' or maybe 'href' or something?

Either that or declare it as xml.