- From: Patrick Stickler <patrick.stickler@nokia.com>
- Date: Thu, 17 Feb 2005 10:49:43 +0200
- To: "ext David Orchard" <dorchard@bea.com>
- Cc: <www-tag@w3.org>, <Norman.Walsh@Sun.COM>, <dareo@microsoft.com>, <paul.downey@bt.com>
On Feb 16, 2005, at 23:09, ext David Orchard wrote: > > I think that a namespace owner can specify exactly what they mean a > namespace to mean wrt the names in the space. The owner has at least 3 > options: > > 1. Say the names in the ns are fixed and immutable > 2. Say nothing about the names > 3. Say the names in the ns are mutable. Well, this is more about how a given namespace owner chooses to utilize a namespace, rather than what they mean a namespace to mean. I.e. best to consider a given namespace infinite -- to already include all possible names that could be grounded in that namespace -- and to focus on the models which specify *which* terms are relevant to the application of that model, irregardless of the syntax of the identifiers. Restriction of the right to utilize terms from a given namespace to the owner of the namespace is really a social, not a technical, issue; such that for that infinite space of names, the owner reserves the right to specify how any presently unused terms might be used in the future. But what counts is not which terms are in use at a given time but in which *models* they are used, and *how*, and one cannot determine anything useful simply by focusing on the namespace itself irrespective of those models. If applications are focusing on the semantics of the models by which content is to be interpreted, rather than on the syntax of the identifiers used to identify terms employed by those models, then "changes" to namespaces should be entirely irrelevant. > > In option #3, it would seem a good thing for the ns author to say under > what conditions the names can change, from "backwards compatible" > changes, where the meaning of bc is defined by the ns owner to even > "completely incompatible" changes. > But this is specific to how the terms are used -- i.e. the model, the schema, the ontology -- not the namespace. Yes, the author of a given model *should* specify how changes to the model affect backwards compatibility, and at what point one version of the model becomes incompatible with applications based on a previous version. But that has nothing to do with namespaces. > Option #2 carries the risk that 50% of software will assume #1 and 50% > will assume #3, and then they will be faced with a tough choice when > they want to version. Does the author retrofit option #1 on to v1, and > then lose the benefits of option #3, or does the author retrofit #3 on > to v2, and perhaps breaking software that made assumption of option #1. > > Seems to me yet another case of where the language designer needs to do > up front planning and thinking in order to prevent versioning problems, > sometimes called architecture... Quite. The problem at hand is that too many folks seem to be mistakenly equating namespaces with higher architectural levels. Patrick > Cheers, > Dave > >> -----Original Message----- >> From: www-tag-request@w3.org [mailto:www-tag-request@w3.org] On Behalf > Of >> paul.downey@bt.com >> Sent: Wednesday, February 16, 2005 6:32 AM >> To: dareo@microsoft.com; patrick.stickler@nokia.com; > Norman.Walsh@Sun.COM >> Cc: www-tag@w3.org >> Subject: RE: Significant W3C Confusion over Namespace Meaning and > Policy >> >> >> Dare wrote: >> >>> Then what does the namespace name identify? XML namespaces >>> seem to grow more and more useless by each passing day. >> >> A namespace provides a means of ownership, and doesn't, IMO, >> identify a strict vocabulary frozen at some point in time. >> >> This demarcation still has some real value, not least to prevent >> clashes of names inserted into a combined document from different >> origins. >> >> Paul >> > >
Received on Thursday, 17 February 2005 08:59:28 UTC