From: pauld@cs.washington.edu (Paul Barton-Davis) Subject: Re: X terminals: Suggestion for projects like SLS, MCC, etc. Date: Mon, 14 Jun 1993 19:47:58 GMT
In article <1993Jun14.145054.20767@wixer.bga.com> rhodesia@wixer.bga.com (Felix S. Gallo) writes:
>mark@taylor.uucp (Mark A. Davis) writes:
>>The X server runs on the TERMINAL. You LOAD THIS ONE PROGRAM on the HOST.
>>The terminals TFTP it. You pay for it ONCE and will support as MANY XTERMINALS
>>as you connect to THAT HOST.
>
>I see that you don't understand simple client-server parlance, so I'll go
>ahead and use your terminology.
Mr. Gallo is out-of-his depth here. Mr. Davis is completely correct in
his description.
>Did I mention that the network is nearly unusable
>at the point where X terminals start breaking even with PCs running X386
>locally, and that you'd have to buy another host and several more
>ethernets?
Is that so ? Perhaps you could explain how the CS department at this
University is able to function with 44 X terminals, and overall 251
network devices, attached to a single ethernet (split by a bridge) ?
>Loading Unix on the PCs is probably just as easy as loading Unix on the
>large, expensive host (rational people call it a "server", btw) which
>you never seem to find breath to mention. What's more, once it's done,
>it's done forever. The win keeps winning, though.
You've clearly never managed an X terminal. I do. 68 of 'em. They are
at least two orders of magnitude easier than managing a Unix host.
I've done that too. Some big (like a 56 node KSR1, or a 20 node
Symmetry) and some small (like Sun-3's and VAXstations (in times
past). Claiming the equivalence of the management of an X terminal
with a machine that runs a configurable Unix system is so
mind-numbingly dumb as to be beyond words.
>>You are really wrong. I have admin'ed both, and you are way out
>>of the water making that statement. No terminal comes anywhere near close
>>to the maintenance of a computer, period.
>
>cough.
See your doctor. Sounds like a bad throat to me.
>This box I'm typing in has been up for 380 days. But this is just
>anecdote; I assure you that properly configured PCs last for a very
>long time, thanks very much.
I see you haven't upgraded the kernel in more than a year. That's too bad.
>And the central host is doubtless able to support 30 X terminals and
>30 users running applications software, multiple windows, and emacs
>for free, right?
On our instructional netwrok, we have (had) 2 hosts supporting 32 X
terminals along with another dozen or so ascii terminals. That makes
about 44 of the *most* demanding users: CS undergrads. Ever seen what
a decent session of nettrek does to your system ? Let alone our own
client/server multi-player dogfight, an application that makes large
database queries look like "grep" in comparison (network load wise,
anyway).
>You're living in a dream world, Mark. The X terminals only provide
>the interface to the user. The cost of replacing disks and the cost
>of CPU horsepower is just as large for the X user as for the PC+X user,
>except the PC+X user has a CPU all to his or her self.
Except that in the host+xterm scenario, there is only *one* set of
disks, *one* cpu likely to be upgrade, and *one* copy of the software
(both host and terminal software included). This is no small thing.
And if you believe that a PC user needs all those cycles to
themselves, well, you clearly don't know too much about workstation
loading patterns.
>I take the time to pursue the optimal. Sue me. :)
Sounds like you could use more time.
-- paul
-- Nonviolence is not a land of milk and honey. Nonviolence is tough. You don't practice nonviolence by attending conferences, you practice it on picket lines, and that can he hard when you are faced with people who may be ready to attack you with rocks, baseball bats, knives, even guns. [ Cesar Chavez ]