On 14 Jun 2001, Deven Phillips <[EMAIL PROTECTED]> wrote:
> Martin,
>
> Loks like the way to do this so that it is most protable and flexible
> is to wrap the current rsync command with alibrary based on libexpect.
> This will allow the library to take password/username arguments and pass
> them along in the the case the SSH is the transport. For future
> developments, we'll be looking to provide some input on the NFS4 version
> of rsync. If I write it correctly, the library will be flexible enough
> to be used by any program that wants to read or write files. Perhaps it
> will even be possible to write a Windows GUI based on this library as
> well (for those unenlightened souls out there).
Andrew and I talked about this yesterday.
Personally I think you can probably get away without needing to
simulate interactive input to rsync: for rsync-daemon passwords you
can send the password in through an environment variable
$RSYNC_PASSWORD, or through --password-file.
It looks like there is no equivalent for OpenSSH, but in that case
it's almost always better to use an SSH key. Certainly some CVS front
ends behaved the same way last time I looked (emacs pcl-cvs, WinCVS,
...)
I guess it depends on how much hassle it is to use libexpect vs plain
popen(). Does it work on Windows? From my experience, tty-style
interaction is pretty broken there, and pipes are about the most you
can expect.
Either way, it would be great to have a foundation for Linux and
Windows GUIs.
--
Martin
PGP signature
