Re: [Gnu-arch-users] Re: multiple committers, again

[Sergio Gelato]

* Martin Langhoff [2004-08-26 12:51:31 +1200]:
> afaik "local" chmod/chown are supported. Remote chmod/chown would
> require support in OpenSSH's sft-server. Sadly, there are several
> threads in openssh's dev mailing list discussing why they won't
> support remote chmod/chown.

It works in OpenSSH 3.8p1. See:

sftp> ls -l 37197796.tar.Z
-rw-r--r--    0 350      200         62431 Aug 26 10:17 37197796.tar.Z
sftp> chmod 600 37197796.tar.Z
Changing mode on /home/gelato/37197796.tar.Z
sftp> ls -l 37197796.tar.Z
-rw-------    0 350      200         62431 Aug 26 10:17 37197796.tar.Z

Also, I'm subscribed to openssh-unix-dev and haven't seen any discussion
of this topic in recent months. There may well have been such threads in
the past, but they are apparently no longer current.

That said, since this is really an issue of access control semantics of
the remote archive, it should probably be dealt with by a helper process
at the remote end. Which would mean something like an arch server. It's
a nice feature of arch that most users won't ever need such a server, 
but still there seems to be a niche for it here.

I was really replying to Stefan Monnier's remark, that the umask rules
(or, more abstractly, the ACL) should be stored in the archive itself.
Since the format of the archive is defined by arch, that would then place
the burden of enforcing the rules on arch.

One can also disagree with Stefan, and put the burden on the arch server
(in an abstract sense: an sftp/webDAV/etc. server used to access archives 
would qualify as an arch server for this purpose) to control access.
Then the answer to rwa would be: "this falls outside the scope of arch; 
talk to your server admins". Which seems to be the majority view here.