[
Lists Home |
Date Index |
Thread Index
]
From: "Eric van der Vlist" <vdv@dyomedea.com>
> > Seems like the API would be more useful, better suited to regular models
and
> > more universally applicable if it returned a set of validated types.
>
> Exactly. Another option is to check if the schema is ambiguous before
> running MSV and that's where RelaxMeter [1] can play a role (even if it
> has its own definition of what's ambiguous): if you schema passes
> RelaxMeter then you can be sure that it's safe to rely on MSV's type
> assignments.
>
> [1] http://www.kohsuke.org/relaxng/meter/
Yes, and has better odds of being translatable to WXS, etc. But this, too,
assumes that the inability to assign a single type to an element or
attribute is a problem that must be fixed. If a string can be validly
resolved to more than one type, this information may be useful to an
application (or to different applications processing the same data).
Throwing away that information seems a case of trying to make the type
simpler than it actually is.
Bob Foster
|
