- From: Martin J. Dürst <duerst@it.aoyama.ac.jp>
- Date: Tue, 5 Apr 2022 13:52:04 +0900
- To: Guoye Zhang <guoye_zhang@apple.com>, Eric J Bowman <mellowmutt@zoho.com>
- Cc: Julian Reschke <julian.reschke@gmx.de>, ietf-http-wg <ietf-http-wg@w3.org>
On 2022-04-05 10:59, Guoye Zhang wrote: > > >> On Apr 4, 2022, at 6:27 PM, Eric J Bowman <mellowmutt@zoho.com> wrote: >> Doesn't HTTP already provide a uniform interface capable of solving this problem? Regardless of upload method... >> >> Server knows the upload was interrupted if the received content falls short of Content-Length, or the connection timed out, or whatever. >> >> Reconnected client makes HEAD request... >> >> Server responds 206 with Content-Range: {zero to Content-Length} -- indicating an interrupted upload, if the client didn't make a range request. >> >> I don't see any prohibition against responding 206 to a non-range request? >> >> Client now knows how to PATCH the resource. Lather-rinse-repeat until 200 OK on HEAD. >> >> -Eric >> > > > Yes, you can implement something that sort of works by combining various defined methods and header fields, but there are many ambiguities: > > First, how does it uniquely identify a resumable upload? Does the server need to send a unique URL to the client? What about the additional roundtrip time? tus-v2 draft solves it by defining `Upload-Token` which is a cryptographically-random token generated on the client side. > > We’ve also looked at the `Range` header but decided against it due to it providing too much flexibility. Features like multipart ranges are extremely difficult to support on the server. If that's a problem on the server side, servers that can't handle that would just send a simplified Content-Range header (e.g. with only a single range). Regards, Martin. > A single integer in `Upload-Offset` models the capability better and sets the correct expectation. That being said, we can revisit this decision if the work group provides a different perspective. > > Overall, we believe that a tightly-scoped standard would benefit implementors and encourage wide adoption. > > Guoye
Received on Tuesday, 5 April 2022 04:52:24 UTC