[Home] [By Thread] [By Date] [Recent Entries]


Dare Obasanjo scripsit:

> Those that want validation can be satisfied with a "type system" where
> the simple types are either strings or [named] regular expressions that
> restrict the lexical space of a particular string.

Almost, but not quite.  The regexes needed to capture a floating-point
range would be extremely daunting, perhaps not possible.  Ranges and
precisions need their own non-regex representation in the system.

> Those that want type
> augmented infosets want them so that they can perform operations on
> values depening on what types they are. They want to add numbers, sort
> dates, concatenate strings, compare equivalence of values, substitute
> and promote types,   etc. 

I agree, but I despair of getting a proper mechanism (short of a Turing-
complete programming language) for describing, for a novel type, what
its legal operations are.

-- 
A witness cannot give evidence of his           John Cowan
age unless he can remember being born.          jcowan@r...
  --Judge Blagden                               http://www.ccil.org/~cowan

Site Map | Privacy Policy | Terms of Use | Trademarks
Free Stylus Studio XML Training:
W3C Member