🚀 go-pugleaf

In article <9011190912.AA15667@wubios.wustl.edu> david@WUBIOS.WUSTL.EDU (David J. Camp) writes:
>The word "Open System" has usually had an unspoken definition in the
>computer industry.  It has meant a computer that had adequate
>documentation or that allowed applications to migrate to and from other
>platforms.

"Open system" has had quite a few connotations.  For instance, the Apple-II
and IBM-PC are considered open, in contrast to the original Macintosh,
because they have bus slots with published specifications, which allows
other vendors to implement peripheral interfaces for them.

>I noticed in the OSI protocol specification that the term "Open
>System" is given a completely different meaning.  It is defined as any
>system that is capable of communicating with other "Open Systems",
>i.e. one that implements the OSI protocol.

As Gumby pointed out, ISO seemed to have coined the term "open system";
they've been working on the OSI standards since the late 70's, but I don't
recall hearing the term used with your definition until the mid-80's.  I
think that after the ISO work got some publicity the phrase acquired
buzzword status, and was adopted for many uses.

However, there is a fundamental similarity between the uses.  A computer
system (or some aspect of a computer system) is open if it provides a
(perhaps de facto) standard (or at least published) interface, to which
other vendors may "attach" their own products (I quote "attach" because
this definition applies to software as well as hardware).  Thus, a system
conforming to the OSI protocols is open with respect to its networking
facilities -- it may be attached to an ISO network and should interoperate
with the other devices on the network.  Similarly, a system that conforms
to POSIX is an open system, because applications that make use of POSIX
services can be attached to it.

Support of portability is very close to this.  Ideally, an ISO network can
be plugged into any OSI-conformant system, and a POSIX application can be
compiled and run on any POSIX-conformant system.

>This definition will probably acquire legal significance because it is
>part of the formal OSI specification.  It will allow vendors to claim
>that their computers are "Open Systems" when in fact it remains
>difficult or impossible to migrate applications to other systems, or
>thorough documentation is nor provided.

It is somewhat unfortunate that they co-opted such a general-sounding term
and gave it such a limited (not to mention circular -- an open system is
one that supports open systems interconnection) definition.
--
Barry Margolin, Thinking Machines Corp.

barmar@think.com
{uunet,harvard}!think!barmar