Refactor WinRM out of CIFS?

[rveznaver]

I would like to propose a refactor of WinRM out of the CIFS package.
The current implementation is written as if WinRM is just an extension of the CIFS protocol, whereas it is a protocol with a different purpose and logic behind it (i.e. it uses a different authentication model from CIFS). I could try to do a pull request, however I would like to know if the project owners would be on board with it?
Thanks!

P.S.: I believe this is needed to cleanly implement more authentication options like #143.

[hierynomus]

At the moment not entirely yet. The idea is that for a "complete" connection you need two parts, the execution part and the file management part.

In the case of SSH the file-copying is handled by either the SCP or SFTP protocol, whereas the execution part is not configurable (just SSH).

For the Windows world there is also not a 1-cut solution. You need WinRM or Telnet for the execution part, and CIFS for the file-management part of the connection. So they're quite intermingled.

What do you think @vpartington?

[rveznaver]

Ok, if the idea is to have a "complete" connection, then couldn't the mixing and matching of protocols be handled by a lighter wrapper? For example, if you put a SSH server on a Windows machine, you could have CIFS file management with SSH remote execution and a different authentication type.

The tight coupling of the file management and execution protocols makes it quite hard to configure one part independently from the other.

[vpartington]

Actually, the fact that file manipulation and command execution are tied so closely is something that has annoyed me for a long time already. The ability to mix and match both would indeed be nice.

Haven't gotten around to actually implementing this though, but a pull request would be welcome! :-)

Of course we'd then need to figure out how to make this backwards compatible...

[gschueler]

+1 :)