On Mar 18, 2010, at 8:43 AM, Tobias C. Rittweiler wrote:
Terje Norderhaug terje@in-progress.com writes:
On Mar 18, 2010, at 2:06 AM, Tobias C. Rittweiler wrote:
Grepping for :end-of-stream only reveals that use in dispatch-event.
Same for %apply.
Any problem with getting rid of those?
Isn't :end-of-stream intended for gracefully closing a connection from the client?
-- Terje Norderhaug terje@in-progress.com
Presumably you make use of that in the CL swank client you're maintaining. Could you add that one, and maintain it, in Slime's vanilla CVS?
The MCLIDE swank client does implement :end-of-stream but only because it was assumed to be in accordance with the protocol. I don't mind eliminating it.
I have verified that neither swank for clojure nor swank for gambit implements :end-of-stream.
With multiple swank clients and servers, it would be useful if we settle how clients are supposed to end a session:
After the elimination of :end-of-stream, the de-facto protocol is that a client should end a session by closing the connection. A swank server should consider a closed connection in between messages to be the end of a session, and do the required cleanup without reporting it as an error.
-- Terje Norderhaug terje@in-progress.com