> James Clark wrote:
> >
> >I very much like your idea. I think it's extremely clean and elegant. I
> >would suggest a name like <tokenized> (goes well with <mixed>) or maybe
> ><split> (since it does the same job as split() in JavaScript, Perl, C#
etc)
> >or maybe even <list>.
>
> Although I agree that Kawaguchi-san's technique is cool, I think that it
would
> encourage bad markup design (i.e., use whitespace rather than tags).
Lists are
> OK, but more advanced structures bother me. Also, I think that "180 cm"
is less
> useful than "180cm". Hmmm.....
However:
- it's simpler (both to implement and specify) than
oneOrMoreTokens/zeroOrMoreTokens
- it provides some functionality that is legitimate and useful, for example
a list of exactly 4 integers; this is something that XML Schema can do, and
that I expect many people will want, but that we can't support with
oneOrMoreTokens/zeroOneMoreTokens
To me, the increased simplicity is compelling, since simplicity is the main
reason for this whole effort. I think it's the job of a Schema language to
facilitate good markup design, not to prevent bad markup design. The only
effective way to prevent bad markup design is to educate schema designers.
James
