- From: Julian Reschke <julian.reschke@gmx.de>
- Date: Fri, 1 Nov 2024 07:31:58 +0100
- To: Michael Toomim <toomim@gmail.com>, Josh Cohen <joshco@gmail.com>
- Cc: ietf-http-wg@w3.org
On 01.11.2024 02:35, Michael Toomim wrote: > On 10/31/24 3:03 PM, Josh Cohen wrote: >> Perhaps you should use both. Much like a regular web server if we >> have the resource we return it, if we don't have it, but know where it >> is we send a 3xx. If we don't have it and don't have any known >> address for it, then we return a 404. > > The problem is that 404 says "this resource does not exist." That means > it has been deleted. The client and caches should react by deleting the > resource. No, that's not what it means. It could be deleted, or it was never there. Or it has moved elsewhere and the server can't or doesn't want to redirect. Or the server doesn't want to say (such as when that could leak sensitive information). > ... > This is a new case because it enables distributed state. In distributed > systems, you want to avoid coordination. The existing codes assume that > the server is coordinating where copies of the state exist. We want to > enable systems where peers can independently choose how much history to > store. This results in server semantics that we haven't seen before, so > our status codes will have slightly different semantics than we are used > to. > ... I'm pretty sure that that conclusion is not correct. Best regards, Julian
Received on Friday, 1 November 2024 06:32:05 UTC