[slime-devel] Re: Some patches

[Luke Gorrie]

Sven Van Caekenberghe <sven at beta9.be> writes:

> I think someone forgot the loop around the let in create-swank-server
> in swank-openmcl.lisp - without the loop you can only connect once and
> the socket is lost

That was intentional, since with `M-x slime' Emacs will call
create-swank-server each time before connecting. Also partly because
in the "old days" we used a fixed TCP port and to keep it open would
mean only one Lisp could run Swank at a time.

Nowadays it sounds like a good idea to keep the server running,
especially if CREATE-SWANK-SERVER is being called directly instead of
by Emacs via *inferior-lisp*.

> Also for all servers, I would suggest an option to restrict
> connections to localhost only, and even make that the default - in
> openmcl, the test is simply (= (remote-host client-socket) (local-host
> client-socket)) right after the accept

Yeah. This already happens in at least the CMUCL backend (by binding
the socket on the loopback-interface). We should define a proper
interface for CREATE-SWANK-SERVER in swank-backend.lisp that includes
a "loopback only" option and implement it for the other backends.

Hands full just now consing onto TODO :-)

> BTW, I think slime is very cool!

Most glad :-)

Cheers,
Luke