I was wondering if much work had been done on this front, and if
not what it would take to do it. There's a lot already in place.
From what I can tell the pieces would be, assuming a TCP server:
1) add port forwarding to the Tramp method (probably ssh only)
2) generate a server file at the remote
3) set Tramp prefix at the remote
3 is easy enough to do with init_SHELLNAME.sh, e.g. something like
export EMACSCLIENT_TRAMP=/-:${HOSTNAME}:
I'm thinking 1 should be done with `tramp-connection-properties',
but there's a little piece missing. Ideally the server port would
be extracted with something like (cadr (process-contact
server-process)), and then that number goes into an ssh arg "-R
0:127.0.0.1:server-port" to allow dynamic allocation of a port at
the remote end. However the message ssh prints for that allocation
then needs to be captured by Tramp/Emacs so that it can be
provided in the remote server file (or some equivalent).
Thoughts? It would be really nice to have something like this.
Thanks and regards,
- Joel
