Path: utzoo!utgpu!news-server.csri.toronto.edu!clyde.concordia.ca!uunet!jarthur!elroy.jpl.nasa.gov!lll-winken!sol.ctr.columbia.edu!samsung!cs.utexas.edu!rutgers!att!cbnewsh!wolf
From: wolf@cbnewsh.att.com (thomas.wolf)
Newsgroups: comp.sys.atari.st
Subject: Re: Is there an X-Client for the ST?
Message-ID: <1990Aug21.015302.24791@cbnewsh.att.com>
Date: 21 Aug 90 01:53:02 GMT
References: <1990Aug20.163330.17619@agate.berkeley.edu>
Organization: AT&T Bell Laboratories
Lines: 53

From article <1990Aug20.163330.17619@agate.berkeley.edu>, by ericco@stew.ssl.berkeley.edu:
> 
> I noticed in your article that you don't think that an X server on the ST
> would work over a modem.  Why not?  Have you seen the X-terminals that are
> widely available?  They use a Sun4 as the host, and they have
> optimized the transmission lines.

Hm.  The only X-terminal I have seen to work reasonably well over serial
lines were those from GraphOn.  I believe they used some data compression
techniques to get their results.
Most X-terminals work because they're connected to hosts via a high-bandwidth
Ethernet.  Those X-terminals supporting SLIP (serial-line IP) don't seem
to run reasonably fast over regular phone lines (ie. 1200-19200bps),
especially when you have multiple clients running.
> 
> If you look at the MIT distribution of X11, you'd see that it is based
> on a small set of primitives.  These primitives are basically
> rasterops.  Sunview is designed on the same set of primitives.
> However, it is designed to run on Un*x, so it can simple check that
> the device supports PIXRECT primitives.  If so, you can run Sunview on
> it.  I think that a very similiar approach could be used for X11.
> 
> Create a version of X11 that runs on Suns using PIXRECT primitives.
> Then, one would have to write a transmission layer for the modem.  I
> have several ideas on this.  The simplest is to transmit the rasterop
> codes over the modem.  A more complex (but faster?) method may be to
> use an aggressive screen update algorithm for rasterops, like what
> emacs does with text.
> 
> Don't get me wrong, I don't think this would be fast, but I think it
> would work.  The X-terminals certainly work.
> Eric
> ericco@ssl.berkeley.edu

There are several ways of optimizing the transmission of primitives over
slow media.  They all have the same "side-effect":  they eliminate X
Window's greatest plus (in my opinion) - being able to run _any_ remote X
client on a network on your local X Server.  Once you start optimizing (ie.
coming up with your own transmission layer, using non-standard/compressed
events, etc.), you're tied to a single host and/or single set of clients
which "speak" this language.

Mind you, I'm not trying to discourage you from writing an API that is
Xlib compatible and works great over serial lines...I have a Sun at work
and an ST at home and could make use of such a beast :-)

Tom

-- 
+-------------------------------------+ "Stupid" questions are better than
| Thomas Wolf   | (201) 949-8063      | no questions at all. No answer is
| Bell Labs, NJ | wolf@spanky.att.com | better than a stupid one.
+-------------------------------------+