On 07/02/2012 06:02 AM, Marc Feeley wrote:
> I also expect a bidirectional port when doing
>
> (open-file "/dev/tty")
>
> and
>
> (open-tcp-client "google.com:80")
open-file doesn't exist and is meaningless. Open for reading?
Open for writing? Open for whatever I have permissions for
(which we know leads to security and other problems)?
> It is conceivable, but non-trivial, to arrange for these procedures to return
> 2 values: an input port and an output port (some Scheme systems do this), but
> then it becomes bothersome to pass these two values around consistenty
> (unless of course you cons them together, in which case you basically have an
> ad-hoc bidirectional port!).
No. The Right Solution IMNSHO is that (say) open-tcp-client
returns a "connection" or a "socket" object. It's a historical
accident (i.e. a design mistake) for Unix to conflate the types
of file/port and connection, since a connection has two files
*plus* other state.
I.e. a connection is fundamentally a different kind of object than
a port. Trying to conflate them is a design bug.
(Some environments (including Java) don't support bi-directional ports
(except in Java's case RandomAccessFile). This means that mapping Scheme
ports to implementation ports becomes ugly.)
--
--Per Bothner
[email protected] http://per.bothner.com/
_______________________________________________
Scheme-reports mailing list
[email protected]
http://lists.scheme-reports.org/cgi-bin/mailman/listinfo/scheme-reports
