[dsdl-discuss] Re: URL (Re: Datatypes strawman)

Rick

>There is a problem with xsd:anyURI (which I named, b.t.w. in an effort to
get names
that were clear) that URIs are so variable in syntax it is very difficult to
do much
validation on them.

The reasons for considering URI/IRI as a separate datatype is to do with the
use to which these "constrained strings" are put in qualifying DSDL names.
There are, in the RFCs, rules that restrict the set of characters that can
appear. These are sufficient to make it relevant to have a separate set of
checking rules. but if I try to define this in terms of a pattern then the
pattern will change over time. If I make it a primitive then I can simply
refer to the relevant documents without having to redefine the pattern
checking rules. (I'm trying to save work, and bulk in the final spec!)

>XML Schemas moved IDs into two parts: the datatype just tests the tokens,
while the uniqueness checking occurs outside, not as a datatyping issue.
I think you will find that for XML Schemas, the value space of a URL
is the string, not the (resolved-to-absolute) locator.

Not sure whether your mixing two different thing here. IDs and URLs are both
subsets of strings, with different character set constraints. ID uniqueness
checking, like URL resolvability, is outside of datatyping. ID uniqueness
checks fall within Part 2's responsibility. But where does the checking of a
namespace URL against the IETF rules come in? This is why I wanted to
introduce a primitive datatypes for URLs and their replacements, to allow a
level of checking for things where URLs are required.

Martin

--
DSDL members discussion list
To unsubscribe, please send a message with the
command  "unsubscribe" to dsdl-discuss-request@dsdl.org
(mailto:dsdl-discuss-request@dsdl.org?Subject=unsubscribe)