On May 24, 2005, at 20:54, Karl Dubost wrote:

A feed validator doesn't have to use a schema, but it could use it for
a non-normative check ("you may have problems" - or even kicking in
more lengthy checking code only when the schema is violated). This is
up to the validator authors.

Validation is something very precise. It can be validated against a DTD, or against a Schema or another grammar language, etc. At least the "Feed validator" could become a "Feed checker" which develops a heuristic to check if the requirements of the specification are verified. :))) "up to the validator authors" :)

The Feed Validator has always been a "Feed Checker" technically. I complained about the naming terminology when the Feed Validator came out. I don't expect the name to change now.

Even with validators that perform a precise check using a schema, it is important to know what kind of check the validator performs. In the case of Atom, the schema that ships with the specification is a RELAX NG Compact Syntax schema with embedded Schematron. When used with a RELAX NG validator that does not support embedded Schematron (eg. http://hsivonen.iki.fi/validator/), the validator will precisely ignore the embedded Schematron.

A hand-crafted checker can potentially be more useful than a validator that is limited to what is checkable using a given schema language. Case point: DTDs and attribute data types.

--
Henri Sivonen
[EMAIL PROTECTED]
http://hsivonen.iki.fi/

Reply via email to