Re: GGI, EGCS/PGCC, Kernel source

Theodore Y. Ts'o (tytso@MIT.EDU)
Thu, 26 Feb 1998 23:43:07 -0500


From: Jason McMullan <jmcc@pepsi.visus.com>
Date: 26 Feb 1998 15:20:15 GMT

Hmmm... Not a bad idea. I didn't intend to muck with the
TTY line discipline, but this _does_ have some posibilities,
such as:

* (PC) Treat the keyboard/PS2 ports as `funny' serial ports.
* (Mac) ADB simply becomes a collection of serial ports

* All TTY lines can hook into an EvStack, with
the input characters filtered through a
(global/per-device/per-tty/whatever) translation
filter (ie keymap) that outputs Unicode keypresses.

(We're using Unicode16 - if at all possible -
throught the EvStack system)

Err.... I'm not convinced at all this is a good idea to route all TTY
lines through the EvStack system. Recall that some TTY outputs need to
be extremely efficient --- think PPP running at 460kbaud, or even higher
speeds in some cases. Nothing I've seen about EvStack has convinced me
that it would be efficient for this type of thing. It may be find for
keyboard input, but that doesn't mean it'll handle high-volume data
transfer. Let's not make the STREAMS mistakes again, shall we?

I originally architected the tty layers so that you could support
*multiple* serial ports doing 115 kbuad kermit transfers on a 386
running at 40MHz. It would be shame to change Linux so that it requires
a Pentium II to accomplish the same feat.....

- Ted

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu