On Sat, Jan 26, 2002 at 06:52:19PM -0700, Sean M. Burke wrote: > And now I'm beginning to wonder about two problems that occur when a C<< >> > code is empty (corresponding to an XML "<C></C>"). > > Notably, those problems are: > How should C<< >> tokenize? > And: > How should C<< >> tokenize?
[latter tokenization being:] > * a C start-code (consisting of the C<< and all the subsequent whitespace) > * a literal ">>" > > > > I'm tempted to just stipulate that codes with the syntax like C<< ... >> > must not be empty, which pretty much allows the latter tokenziation in both > cases. > > First, there's the completely obvious argument that C<< ... >> codes were > devised specifically to handle the cases where the intended content > contained a literal ">", as on C<< $foo->bar >>, so using them with > no-content is daffy. I think I'd be quite happy with C<< >> being illegal if it contains only whitespace. If someone wants to write C<> or C< > then they can use single <>, surely? Nicholas Clark -- ENOCHOCOLATE http://www.ccl4.org/~nick/CV.html