FWIW, I'll just repeat my previous statement about these error specs. IMO, the issue is that the whole concept of farming out functionality to untrusted parties for tests that need to be trusted (i.e. done by the publishing tool itself), is fundamentally flawed. Nobody will use it.
-1 to both error paces. Also FWIW, in case one does go through, I'd normally suggest just using POST, but if it's the case that Atom resources which already accept POST also need to be able to accept error reports, then due to a deficiency in HTTP (lack of mandatory extensions, and the lack of a standard means for declaring extended intentions), I consider ERR/ERROR a reasonable option. Mark. -- Mark Baker. Ottawa, Ontario, CANADA. http://www.markbaker.ca
