Re: SFTP extensions draft should be submitted as working group item

[denis bider <ietf-ssh%denisbider.com@localhost>]

> Because one must use read and write, it is harder than it
> should be to implement simple transfers where both sides
> are 100% sure that a whole file has been transferred
> (as opposed to part of a file).

That is precisely it, SFTP is NOT a file transfer protocol, it is a remote filesystem protocol, and the SFTP acronym is just a totally misleading name.

People come to us all the time with the preconception that SFTP is FTP over an SSH connection. But it isn't, and I, for one, find this architectural decision welcome. On the other hand, indeed, it makes it a bit more difficult to define what an "upload" and what a "download" is, but I don't consider that a major problem.

I find Jeffrey's frustration understandable, but I think we have potentially the best, the cleanest and most straightforward remote file access protocol on our hands. It requires a few more tweaks to be complete, and I find that all these latest tweaks that were made were necessary. Now we can either specify a few simple tweaks, or botch that and leave those things unspecified, resulting in a standardization void, an implementation chaos and new standardization efforts attempting to fix that in 10 years' time.

I'm fine if OpenSSH doesn't see the need to implement anything other than SFTP version 3. That's totally OK with me, because you know what? It makes it that much easier to compete with them on quality. But I, for one, find the additions of the latest SFTP versions - ACLs, the check-file request that's in the extensions draft, the TEXTMODE flag, the REALPATH behavior change - quite necessary if SFTP is to become a universally useful network protocol. And that is something I want it to be.