- From: Jan Algermissen <algermissen1971@mac.com>
- Date: Tue, 22 May 2007 21:57:35 +0200
- To: noah_mendelsohn@us.ibm.com
- Cc: Mark Baker <distobj@acm.org>, David Orchard <dorchard@bea.com>, Marc de Graauw <marc@marcdegraauw.com>, mark@coactus.com, Marc de Graauw <mdegraau@xs4all.nl>, www-tag@w3.org
On 22.05.2007, at 19:06, noah_mendelsohn@us.ibm.com wrote: > > Mark Baker writes: > >> I'm ok with the second part of that, but if a new version is >> compatible, then a new media type shouldn't be required. > > Which begs the question, what do you mean by compatible? This could be rephrased to "what are the allowed variations between two versions of a format if they are to remain the same media type?". Has that ever been thought through? Jan > Almost surely, a > new version of a language will introduce both new syntax and > corresponding > new information conveyed by that syntax. So, in that sense, there's > almost always some incompatibility when new constructs are used. > > Sometimes the new information is in some sense orthogonal to the old, > allowing one to ask separately which pieces of the content are > "understood". Sometimes, the new rules change the implications of > the old > fields. For example, if the old version says that case doesn't > matter > for a field, and the new version of the language says that case > matters, > then even documents that were legal in the old language may be > taken to > convey information that might not have been intentional when the > document > was authored. Even in the simple case where the new content is > completely orthogonal syntactically and semantically, there is the > question of whether users consider it OK to to ignore if it's not > understood. > > For reaons like these, I don't think using the word "compatible" > without > explanation or qualification is appropriate. I do agree that for some > families of languages, one can define useful notion of > compatibility that > make it appropriate to consider using a single media type for all > of the > versions. I don't agree that those styles of language evolution are > always what people want. > > -------------------------------------- > Noah Mendelsohn > IBM Corporation > One Rogers Street > Cambridge, MA 02142 > 1-617-693-4036 > -------------------------------------- > > > > > > > > > "Mark Baker" <distobj@acm.org> > Sent by: mark@coactus.com > 05/18/2007 08:04 AM > > To: "David Orchard" <dorchard@bea.com> > cc: noah_mendelsohn@us.ibm.com, "Marc de Graauw" > <marc@marcdegraauw.com>, "Marc de Graauw" <mdegraau@xs4all.nl>, > www-tag@w3.org > Subject: Re: Media types and versioning > > > On 5/17/07, David Orchard <dorchard@bea.com> wrote: >> What about incorporating a version # in the media type? >> >> Say application/soap+1dot2+xml, etc. > > That's fine. On the ietf-types list, I think I've recommended > something similar for at least one format which made backwards > incompatible changes. > >> Then say in the media type definition that by definition any minor >> version change is compatible, and any incompatible change will be >> accompanied by a major version change. > > I'm ok with the second part of that, but if a new version is > compatible, then a new media type shouldn't be required. > > Mark. > > >
Received on Tuesday, 22 May 2007 19:58:30 UTC