Community
Participate
Working Groups
In the "Connecting to a remote Windows server" help page says "Use of the server daemon on Windows systems is not recommended.". So I tried to run the server directly. The instruction says: "You will then have to enter this port number in the port property for the Files subsystem for your connection in the Remote System Explorer (see Connecting below)." Below it says: "To check your port number, right-click your connection or subsystem from the Remote Systems view and select Properties. Click Subsystem to see the relevant information. If your port is "0," then your Remote System Explorer communications server will pick any free port on the Windows server. If you specified a port number when starting the server, you need to enter it here, for example, to work with a firewall." This simply makes no sense to me. Maybe I'm to stupid, but I was not able to get it running with a server. See: http://scharf.gr/eclipse/rse/problems/windows-server/
I agree that it's currently too hard to connect to a running server. I think the right solution would be to be able and enter the running server's port directly in the wizard (and store it with the connector service). That's what bug 142969 requests.
For non-windows, I would expect that connecting to a running server is more of a developer testing scenario rather than a normal user scenario (such as the case where a daemon is used). Since it's pain to have to manually start a server, I would expect that approach to get tiresome very quickly. The only real reason it's exposed today is for debugging and testing situations where the daemon gets in the way. As such, I don't really think we are really serving the user in trying to enhance the user experience in this particular case. However, as use of the daemon is not recommended on windows, I understand why the running server approach ends up in play. The ultimate solution would be to have a proper windows service that acts as a daemon. I'm not familiar with how to write these services - do any of you have experience in that? As Martin pointed out, exposing the server port in the new connection wizard should be handled in 142969.
I think there is another important scenario where a "running server" is needed: User has no root access on remote, daemon not running, and rlogin is forbidden --> user starts the remote server through ssh, optionally creates an ssh tunnel and connects to the server This is even documented in the user docs. Provided that the same server port is used each time, it can all done by scripts and is therefore not tiresome at all. More generally, any type of server launching that we do not have a specialized Server Launcher for yet, is a candidate for needing an "already running" server. What we should probly provide is a "generic server launcher" that invokes user-contributed scripts to actually do the server launching on behalf of the user, and reports back the port to connect through e.g. stdout. For the sake of a hyperlink, the port to be used is exposed in bug 142969.