- From: Julian Reschke <julian.reschke@gmx.de>
- Date: Sat, 23 Aug 2008 13:07:33 +0200
Ian Hickson wrote: > ... > Put it this way: > > Using URIs is definitely a problem; people invent entire declaration > syntaxes to route around them (c.f. XML namespaces). > ... I do not agree that this is a problem. > Using unambiguous class names like Java does is maybe a problem due to > clashes. > > Unless we can demonstrate that the clashes are more of a problem than the > use of URIs, we would be foolish to put our eggs in the URI basket. We would be foolish not to, because URIs are already the established naming system on the web. Introducing another one requires mapping between both, which will be expensive and fragile. Examples: - RDF predicates - WebDAV properties, condition codes, report names - XML namespace names - XSLT extensions - XML Schema datatypes (yes, also used in Relax NG) - XHTML modularization - JCR node names, property types and node types There are a few examples of naming systems that I use in my daily work. I'm sure if you ask other people, they'll quickly come up with other examples. > One good way to see whether it really is a problem is to look at the Java > ecosystem. Have clashes been a big problem there? Classes that were put into the "default" package certainly were a problem -- when those need to be re-used, they usually have to be renamed (into a package). That may be no problem for a few Java classes, but more of a problem with names that are widely used outside some Java code. Not using URIs and inventing a new naming scheme instead means working *against* existing naming schemes. So I think it's up to you to prove that it's worth the cost. BR, Julian
Received on Saturday, 23 August 2008 04:07:33 UTC