Re: [ebxml-cppa] Versioning for CPP: new proposal to drop versionattribute on CPP.

[Christopher Ferris <chris.ferris@sun.com>]

My recollection was the the version attribute applied to
the CPP schema version, not the instance of a given CPP
that was somehow modified. It has the same semantics
as the version attribute in XSLT[1].

Cheers,

Chris

[1] http://www.w3.org/tr/xslt

Dale Moberg wrote:

> Hi,
> 
> I am stumped on why we have versioning for a CPP.
> 
> The cppid is to supply a unique identifier for a CPP.
> For the sake of argument, let's say that this means
> that no two CPPs can have the same cppid. 
> 
> When are CPPs the same? For the sake of argument, let us say
> that CPPs are the same when their hash values, for
> the purpose of signing, are the same. Conversely, we
> have two CPPs when their hash values differ.
> 
> So if we changed the version _value_, and left all else
> the same in a CPP, we would have two CPPs with different
> hash values and the same cppid. This contradicts
> our assumption that no two CPPs have the same cppid.
> 
> Therefore, we cannot really have _two_ CPPs that differ
> only by version number (on the previous assumptions).
> 
> So we need to say what it means for CPPs to differ
> enough to be different versions, but not differ
> enough to warrant having different cppids. I think
> this is either a rabbit hole or a rat hole, and
> I now think that either
> way we should avoid entering therein.
> 
> I am now inclined to say we should drop the version attribute
> on the CollaborationProtocolProfile element. We should also
> just say that the cppid value should be a globally unique
> identifier. 
> 
> Should we say anything about format of that guid?
> Should we reference some other standard that has 
> solved the issue of how collisions
> in creation of this identifier are avoided? 
> If so, does anyone know what standard would be good
> to reference here?
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> ----------------------------------------------------------------
> To subscribe or unsubscribe from this elist use the subscription
> manager: <http://lists.oasis-open.org/ob/adm.pl>
>