Path: utzoo!mnetor!uunet!husc6!mit-eddie!uw-beaver!uw-june!context
From: context@june.cs.washington.edu (Ronald Blanford)
Newsgroups: comp.sys.ibm.pc
Subject: Re: hooking PC's together
Message-ID: <4368@june.cs.washington.edu>
Date: 7 Mar 88 17:41:52 GMT
References: <673@pasteur.Berkeley.Edu> <63200030@convexe> <1305@qetzal.UUCP> <3438@cup.portal.com> <1894@uwspan.UUCP> <730@gethen.UUCP>
Reply-To: context@june.cs.washington.edu (Ronald Blanford)
Organization: U of Washington, Computer Science, Seattle
Lines: 24

From uw-beaver!mit-eddie!ll-xn!ames!pasteur!ucbvax!unisoft!gethen!farren Mon Mar  7 09:23:06 PST 1988

In article <1894@uwspan.UUCP> root@uwspan.UUCP (John Plocher) writes:
>Because the parallel adapter in the PC and AT is UNI-Directional - it can
>send but it can not receive data.

In article  <730@gethen.UUCP> farren@gethen.UUCP (Michael J. Farren) writes:
>there is
>an input data buffer on those data lines as well as an output buffer.
>You have to switch a bit or two on one of the 8255's to get data in,

At the UW Computer Science Department we have successfully used the
parallel port for bidirectional data transfer between a PC and a
prototype special-purpose image coprocessor.  It is fast and the
interface at the other end is much simpler to implement than a serial
connection. 

I suspect this technique is not used more extensively for precisely
the reason Michael gives: you have to choose and explicitly switch
between input and output.  Half-duplex serial connections have been
anathema outside the IBM mainframe community, so designers tend to
avoid them even when the problem may require nothing more.

-- Ron