Empty lists?

Daniel W. Connolly (connolly@hal.com)
Mon, 14 Nov 94 16:11:10 EST

Dan> * (LI)+ => (LI)* in OL and UL: why?

eric> The group seemed to agree that the degenerate case should be allowed.

lee> so that an automatically generated list can be empty; someone in
lee> Chicago [who? where are the minutes when you need them? :-)] argued
lee> v. strongly for this, and I think it's reasonable.

Grrr... well, that much I can infer from the comments on the page.
What I cannot deduce is why? In the prose of the spec, you pretty much
have to use consensus to decide arbitrary issues like wording and
organization. But in the DTD, can we not make technical, or at least
emperical arguments to motivate changes?

My argument for (LI)+ is that (1) it agrees with the data that I have
examined, and (2) it catches editing errors. If I type <ul></ul>, I
probably didn't mean it. I probably had some items that I was going to
fill in later and forgot.

ehood> I have a ToC program that generates nested lists automatically.
ehood> Conditions can exist when the following construct can occur...
ehood>
ehood> I guess I'm asking what is the best way to solve this problem. This
ehood> problem is different than (LI)* since markup is being generated where
ehood> UL is a direct child element of UL. The following

I'm quite confident that you can change your ToC program to generate
valid code.

ehood> Another(?) way to state the problem is:
ehood>
ehood> Should one be able to have a N+2, or greater, level list be a
ehood> child of a Nth level list?

Why? Can you give a more concrete example where this sort of thing
is motivated?

ehood> HTML already allows similiar constructs by allowing Hn elements to
ehood> occur in any order (implying arbitrary levels of sections). Should the
ehood> same leniency be applied to other elements as well?

I think nested levels are quite different from numbered headings. In
any case, I think we would all want Hn elements to obey strict
ordering, if we were designing from scratch. But there's a lot of
existing data out there. As far as I can tell, nobody uses empty
ULs. Evidence to the contrary is welcome.

Dan