"Seaparating REST Facts from Fallacies"
- Steve Bjorg wrote:
> On Dec 3, 2008, at 5:07 AM, Bill de hOra wrote:I think so - if compliant code can't process the new version, either
>> Steve Bjorg wrote:
>>> On Nov 24, 2008, at 11:37 AM, Subbu Allamaraju wrote:
>>> > This treats "extensions" are required elements, and that is a
>>> > dangerous approach. Extensions are "extensions".
>>> Yes and removing the extension would still make it a valid Atom
>>> entry. However, the specific application that you might submit it to
>>> would reject it, because it requires the presence of the extension,
>>> which feels like an acceptable compromise.
>> I wouldn't recommend that approach. It makes systems unpredictable and
>> needlessly complicated (forced upgrades on clients or legacy switching
>> on servers). I think you'll also find it'll be hard to allocate bugs
>> between clients and servers - IME formats that allow retroactive
>> mustUnderstand cause social problems more than technical ones.
> If someone were to rely on media-types to capture these constraints and
> the application evolves, should the media-type be versioned?
version the media type (via a type param) or mint a new media type.