Dissenting argument on namespace URI

From
Robin Cover
Date
2007-04-17T14:54:00+00:00
ID
Thread
Dissenting argument on namespace URI
Short answer: some legacy NS URIs do violate the rule; these cases were
grandfathered.

-----------

Longer answer:
> missing something obvious?

No, you caught the obvious. In the 2005-2006 timeframe, several
activities in TAB and Staff (with two OASIS member reviews) tried
to address issues relating to namespace design.  At the time,
no direct support was available for adjustment of server config files
(e.g., for use of mod_rewrite directives), so we only had a crude
form of "redirect" to work with as we tried to manage the TCs'
namespace URIs and resources.

Over time, it became clearer to TAB members, Staff, and other
experts that overloading at the point of a directory/NS URI
would create problems -- given our commitment to transparently
display all directory contents.  In the interim, several TCs
had created content in directory URIs which collided with
namespace URIs, so we had to use clumsy, non-predictable
workarounds to provide views of the directory contents (index).

The result of this experience and many design discussions was to
create the rule you now see at "TCs should avoid creating
collision/confusion and semantic overloading..."

We grandfathered the instances on the file system of docs.oasis-open.org,
and will probably allow TCs that started out with the (now) deprecated
pattern to play out the remainder of their TC work in a consistent
way.  For new design work in TCs that are minting NS URIs, we intend
to require adherence to the rule "TCs should avoid creating
collision/confusion..."

Robin

On Tue, 17 Apr 2007, Anthony B. Coates (Miley Watts) wrote:
> Thanks, Robin.  Maybe you could help me in understanding the OASIS
> namespace URI guidelines
>
> http://docs.oasis-open.org/specGuidelines/namingGuidelines/resourceNamingV07.html#NamespaceDesign
>
> a little better.  In particular the section
>
> ----
> TCs should avoid creating collision/confusion and semantic overloading at
> the point of a XML Namespace URI which could be mistaken for a regular
> directory URI, or vice versa; thus, for any Type 1 or Type 3 HTTP scheme
> namespace URI terminating in foo or foo/ (e.g.,
> http://http://docs.oasis-open.org/tcName/path/foo), no URIs should be
> created for resources matching "*" in
> http://http://docs.oasis-open.org/tcName/path/foo/*. IOW: we don't put any
> content in directory foo/.
> ----
>
> I understood this to mean that if the genericode 1.0 namespace URI is
>
> http://docs.oasis-open.org/codelist/genericode/1.0/
>
> that we can't have any documents whose access URI is of the form
>
> http://docs.oasis-open.org/codelist/genericode/1.0/*
>
> However, this is not consistent with the first two of the RDDL examples
> that you posted
>
> http://docs.oasis-open.org/ws-rx/wsmc/200702
> http://docs.oasis-open.org/ws-rx/wsrm/200702
>
> where there are documents in the directory of the namespace URI.  Am I
> missing something obvious?  Thanks,
>
> Cheers, Tony.
>
> On Mon, 16 Apr 2007 21:36:16 +0100, Robin Cover