Re: [syndication] OCS Version 0.5 draft

[Morten Frederiksen <mof-syndication@mfd-consult.dk>]

Hi (Ian),

On Wednesday 04 September 2002 20:05, you wrote:
> > I don't know if versioning the modules is needed...
> I disagree. When a module is finalized then it should be at version 1
> and that's the version people should use in live systems. Subsequent
> versions could subtly alter the semantics of any module elements.
> Maybe a future version of the alias module could add some sort of
> pattern matching attribute.
As long as the change is something like this, I see no reason to break 
existing implementations by changing the namespace, if the absence of the 
element implies the same meaning as the previous "version".

If the module is up-versioned for a new optional element or an optional 
change to the meaning of an existing element, every implementation out there 
will no longer work at all, since it would need to at least update the 
namespace. I think this is the same issue that resulted in "XML 1.0, 2nd 
edition".

> An agent dealing with the module needs to
> hard-code behaviour and that's easier to do if every element's
> semantics are guaranteed through the use of a unique identifier.
Yes, that's why I think that substantial changes to existing elements, like 
changing the meaning, needs to lead to a new namespace, but that new optional 
elements don't.

As an example, the RSS 1.0 mod_content module is standing to gain a new 
element, but there are no changes to the existing, and as such the namespace 
doesn't need to change.

> It's up to individual agents to equate behaviour for elements from
> different namespaces, much like the old title vs dc:title debate.
True, but if the namespace is changed every time there's a small addition or 
change, you'll have developers dropping out, refusing to try to keep up.

> > Also, in contrast to the description in section 4, the example doesn't
> > list the aliases in an rdf:Alt container element. Keeping with the other
> > aspects, I can imagine that the decription is OK, but the example should
> > be fixed.
> In this case the example is correct - I'll change the text to reflect
> this.
Hmm, so for the core OCS, channels are listed in a Bag element and formats in 
an Alt element, but the alias list is not?
This may lead to confusion, but since your example clearly indicate that 
multiple aliases can be present, I guess it will work, it just seems 
inconsistent.

One other thing I noticed just now: The only occurences of rdf:Description I 
see in OCS 0.5, as compared to OCS 0.4, is the ones for each alternative 
format/language/schedule. Would it be possible to get rid of those complete, 
to complete the migration to an RSS 1.0-like style?


Regards,
Morten Frederiksen