Cool! :-)
I have not been following this discussion for more than a few months, as I have been really busy with Atom-OWL. I thought it had died out completely.
I am +1 on something like PaceServiceError, with the idea of turning it perhaps into an Internet Draft if that is possible. It seems to me to be a very generic HTTP addition that could well be of use all over the web.
I would prefer the method to be called ERR or ERROR over GRUMBLE, because it is well understood in software engineering that functions can return error codes.
Also I prefer the simplicity of ERRing the resource itself, and nothing else. That removes the need for an extra header for every response and really simplifies the mechanism of reporting the error: The client, in its error report, just needs to send back to the resource the exact same request it used in its GET, POST, PUT or DELETE, +perhaps some extra information such as the time of the original request and the method of the original request.
I see that Atom could be fixing a very big problem on the web by demonstrating how this is done, especially if it does it in such a way that it is completely generic, and so easily added to every other web app out there.
Henry Story
On 4 Nov 2004, at 20:52, Tim Bray wrote:
On Nov 4, 2004, at 10:10 AM, Robert Sayre wrote:
I'm -1,000,000 on using POST for this feature. Too bad, but since Atom is so open I'm free to try it without the approval of this WG.
Well, on reflection, it seems to me that, while I still don't believe that GRUMBLE really brings much of a security pay-off, on reflection it doesn't add any significant costs, so I can live with PaceServiceError as written.
And I do think it would be a net plus for Atom to have a non-compulsory error-reporting built-in. -Tim
