From: Bill Stewart <stewarts@ix.netcom.com>
Date: Mon, 15 Apr 1996 09:12:57 +0800
To: Duncan Frissell <frissell@panix.com>
Subject: Re: Open Systems, Closed Systems, & Killer Apps
Message-ID: <199604142149.OAA28361@toad.com>
MIME-Version: 1.0
Content-Type: text/plain


At 10:14 AM 4/10/96 -0400, you wrote:
>-----BEGIN PGP SIGNED MESSAGE-----
>
>Various correspondents have pointed out that X.25 is an "open system" in
>that it is not proprietary.  I knew that.  I was thinking more of
>hierarchical vs peer-to-peer.  I have been under the impression that
>TCP/IP connections are more peer-to-peer between different sorts of
>networks (or nodes) than X.25.  Isn't X.25 more of a standard for a
>single network?  Don't X.25 networks need someone more "in charge" than
>TCP/IP networks, or am I mixing up different layers on the OSI reference
>model?  

X.25 is an interface between a Data Terminal Equipment and a Data Communications
Equipment, rather than a whole-network format like IP and TCP.
X.25 networks often have random proprietary internals, and they're designed
for a world where there IS only one network, because after all, there's only
one Phone Company; the X.75 protocol lets X.25 networks talk to each other.
But if you look at the higher levels of the protocols, they're not really
that different than TCP applications - you've typically got a listener 
application waiting around for connections from client programs.
It feels a bit less peer-to-peer because usually the service you want
is the X.3/X.28/X.29 stuff that's X.25's equivalent to telnet,
so the server end is a MainFrame, and the client end is a terminal pad
that you've connected your 3270 or dumb terminal to.  But you can do other 
things as well, if your computer environment will support it.

Simon Spero wrote:
} you can call X.25 a lot of things, but proprietary is not one of them. 
} X.25 did not fail because it wasn't open; X.25 failed because it was crap

It's not dead yet, and you can't even say it's failed, given that it's still
in wide use in much of the world.  X.25 was design to work on networks with
really bad bit loss - we're talking modems on barbed wire here, or whatever
the French use instead of barbed wire, and in days when computers were _slow_.
Yes, it's bureaucratically designed, and parts of it genuinely are ugly,
and it does lots of work at Layer 2 that these days you'd do at a higher layer.
And, yes, it's a lousy environment to do full-duplex character echo over.
But it works ok for a large fraction of the world's data communications,
which are designed for less interactive environments.  It's fine for email.
It's fine for 3270 fill-in-the-blanks applications.  It's fine for pre-Web
online service applications like CompuServe.  It reeks badly
for client-server applications which do a dozen little handshakes per
transaction,
which are designed assuming they're on a LAN and fail badly when stretched
across an ocean, but you'll find they often do badly on frame relay as well.

Are the Internet and Frame Relay both better ways to do anything than X.25?
Yeah.
Would I want to do interactive work on it?  Of course not.  Would I like to do
another project getting vendors to modify their X.25 and CLNP to support a set
of seldom-used options that some security consultant once convinced one of my
customers they needed?  No way.  Was I susprised that not only does AT&T still
use X.25 in some of its older dialup networks, but that it's still very big
overseas?
Well, yeah :-)
#					Thanks;  Bill
# Bill Stewart, stewarts@ix.netcom.com, +1-415-442-2215