Lists Home |
Date Index |
Elliotte Rusty Harold wrote:
> At 1:11 PM -0700 10/24/03, Joe English wrote:
> >Now with sufficient intuition and guesswork, you can _probably_
> >write software that _usually_ works _most_ of the time if all
> >you have by way of a specification is a collection of sample
> >inputs. Personally, I'd rather have a schema (RNC format,
> >if you please) and skip the guesswork.
> And then what do you do when the document doesn't adhere to the
Fix the bug in the software that produced it, or fix the
document itself if it's human-authored, or amend the schema
to account for new structures (and double-check any programs
that process inputs of that type to make sure they can handle
the more liberal schema).
> The schema's a start, but not an end. Robust software does
> not assume the schema is actually followed.
True, but _correct_ software is free to make that assumption :-)
It can blame some other part of the system if it's violated.