geronimo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Bill de hÓra <b...@dehora.net>
Subject Re: Version numbers in namespaces
Date Sat, 22 May 2004 14:41:39 GMT
Alan D. Cabrera wrote:
> Interesting.  Personally, I don't care for the #info aspect of the
> namespace.

Some background on #:

In RDF-land (where this idiom comes from), this is considered by a 
number of people as broken and I imagine it's quietly being 
deprecated over the last few years. It's good that people want to 
make readable URIs for the purposes of quick eyeballing, but the 
idea of algorithimically hacking into an opaque name by putting 
markers into it is dubious. It doesn't make things clearer that the 
strings we're using for XML Namespace /names/ are in fact 
/structures/ wrt the Web (but that's a long running and heated 
perma-thread in XML).


Personally, I don't encourage #fragid use in namespaces - you don't 
need fragment structure for (an opaque) name. Again personally, I 
would disagree with the Cocoon consensus. Mixing XML namespaces and 
document structures is a messy affair.

Beyond that, current practice seems to be focusing on the idea of 
putting time into names, not version numbers. So you have something 
like this:

   http://www.apache.org/foo/bar/2004/05/22
   http://www.apache.org/foo/bar/2004/05
   http://www.apache.org/foo/bar/2004

That's useful insofar as it gives flexbility as to how to decide to 
upgrade physical schema wrt a version number. If you use version 
numbers you tie your names into the upgrade strategy of whatever 
versioning policies are in place. Possibly over time, that will not 
be what you want, given that versioning is just about a difficult 
problem as your are likely to find in this industry (it's right up 
there with naming and cache invalidation in my book).

For example, I do a lot of work with particular XML enveloping 
structure that has embedded "1.0" in its namespace. I dearly wish a 
date had been use as would have avoided interminable confusion and 
deadlock around what were non-breaking and editorial alterations to 
schema. "2001" would have been easier all round.

If the consensus here is to put version numbers into namespace names 
then I suggest that only the major number is inserted. I also 
suggest that a versioning policy be laid down first. But with dates 
you can defer such policy decisions and/or change them if required.

cheers
Bill de hÓra




Mime
View raw message