Windows clients can't mount a remote tunneled server since we can't tell them what port to use (sigh). So maybe we should always do a reverse mount? (server mounts shares from the client)
We need to find out the list of shares (pysamba?) and allow the user to configure what is mounted (read-only, password,...)

The actual mounting is the easy part:

mount.cifs //127.0.0.1/public ./guest -o port=21062 -o guest

But we have to make sure we always unmount stuff when the server goes down... Problem is that some files might still be open!

We should also deal with non-tunneled setups nicely: simply connect to SMB on client or server directly. This weighs in favour of supporting client-mounts-server option.

We can figure out what shares are available with (needs a bit of magic parsing... sigh):

smbclient -U guest -N -L localhost

or

smbclient -A /path/to/auth-file -L localhost

​pysamba does not seem to help here, unless I am missing something.

See bug #91 for the next step.

In ssh-mode, we launch a remote X command and putty is still running when the command terminates.
This causes the session to remain "active" on the client despite the fact that it is really finished...

This happens with all the clients we can use: TortoisePlink.exe, Plink.exe as well as openssh (via cygwin). Looks like the clients refuse to shutdown whilst there is still a tunnel (the X display connection) so we may have to use a trick on the server and run the command via a wrapper which will take care of killing the ssh process when the command terminates.

Some apps only supply an SVG icon (ie: wine notepad).
Either we convert those on the fly to another format (how?) and cache this on disk.
Or we get Windows/gtk to play nice and load the svg pixbuf loader...