[sc34wg3] CTM should be context free

[Robert Barta]

On Mon, Jan 28, 2008 at 02:54:38PM +1300, Xuân Baldauf wrote:

> > 	donald = mailto:donald at duck.org {
> > 	  - "Donald Duck";
> > 	  - "Anders And" @norwegian;
> > 	}
> >
> > This would work, unless you had declared a prefix called
> > 'mailto', in which case the parser would try to interpret the
> > IRI as a QName and you would get a syntax error.

> This is neither regular nor context free. IMNSHO, CTM should be a
> context free language. We should not make CTM as hard as C++ in parsing,
> both for humans and computers. I also do not see necessity for CTM being
> not context free, does anybody else see such a necessity?

As implementor I would not so much worry about the
context-sensitiveness.  That can be handled efficiently. But the
conceptual picture a user should be as context-free as possible.

Even the template thing (as useful as it is) and the prefix thing (as
useful as it is), is already stretching that to the limit.

\rho