🚀 go-pugleaf

In article <1jp3mmINNnpo@cs.widener.edu>
	jefferis@manas.engr.widener.edu (Raymond P Jefferis) writes:
>In article <1993Jan21.160050.4997@tse.uucp>
>	markt@tse.uucp (Mark Turner) asks:
>> Is it possible to hang X terminals off a NeXT server and use the native  
>> NeXT desktop from the terminals?
>> Comments .... ?
>
>What is a "NeXT server?

It's someone like myself who's purpose in life is to make my NeXT
as comfortable as possible, including slavish worship at the keyboard
all hours of the day (and night), and purchase of silly peripherals
such as CD-ROM drives.

Later,
--
scott hess <shess@ssesco.com>                        <To the BatCube, Robin>
12901 Upton Avenue South, #326  Burnsville, MN 55337 (612) 895-1208 Anytime!

In article <1993Jan21.160050.4997@tse.uucp> markt@tse.uucp (Mark Turner)  
asks:
> Is it possible to hang X terminals off a NeXT server and use the native  
> NeXT desktop from the terminals?
> Comments .... ?
What is a "NeXT server?

Mark Turner writes

> I've come to the conclusion that a "CPU on every desk" is not the
> right solution for many businesses. As soon as a $20,000 workstation
> lands on a corporate desk, it needs more memory and more hard disk
> ... and the CPU itself it out-of-date and too slow in just a few
> months.

> The X terminals can be replaced, or upgraded, without much fuss
> and little expense.  And it's simply more efficient to do system
> administration, and manage most of the hardware problems, centrally.

Your question is making the implicit assumption that the above is
true.  Having watched X terminals in action, and seeing what they
do to a network, I don't believe that it's a valid assumption.

X terminals aren't any cheaper than a well-equipped 486, which can
then run Unix and X native, with lots of disk, and not swap over
the network.  A full Unix machine can also be easier to administer
in a complex, shifting environment.  (Not easy, mind you, just
easier than a X terminal.) Especially if using NetInfo.

Also, CPUs are a cheap resource compared to networks.  This will
remain true for at least 5-10 years.  And X terminals simply _swamp_
the network.  It just isn't effective to be running more than a
few on a single subnet.  And for responsiveness, you need an adequate
amount of peak computing power.  That ends up being just about one
CPU per terminal in the end.  That's in addition to the high-power
processor in the X terminal.

X terminals might still be more effective if it were easier to
upgrade central CPUs than distributed CPUs.  I don't believe this
is true.  Upgrading a distributed CPU is a fairly simple task, and
doesn't take a whole subnet down during the upgrade.  Most people
will never even need an upgrade if provided with adequate power in
the first place.

X terminals strike me mostly as "penny wise, pound foolish".  Like
mainframes, only much more so.

> I often hear NeXT advocates complaining about the X protocol, and
> promoting Display PostScript ... possibly with good reason. Is it
> possible to hang X terminals off a NeXT server and use the native
> NeXT desktop from the terminals? Is there any such thing as a
> PostScript Terminal? If not, why not? Is anyone out there adopting
> this sort of strategy?

Even given the above, it's still an interesting question.  It is
true that a well-written Display PostScript protocol could end up
being more efficient than the X protocol, but applications would
need to be modified to use it properly.  You would want to be able
to download event handlers into the server, and only send back
events to the application client only when necessary.  Note that
this is a different strategy than you would want to use when the
server and client are on the same machine - event handling requires
math, and is more efficiently done in C, so when there is no
transmission cost for events, all of them should be sent to the
client and handled there.

If the application's drawing code was written well, the amount of
network traffic to draw using Display PostScript ends up being much
smaller than if done using the X protocol.  An application talking
Display PostScript to an X terminal with the Display PostScript
extension, for example, would not swamp the network as much.  NCD
has announced such an X terminal, so in a way "PostScript terminals"
will exist.

However, a terminal market for a specific protocol would only make
sense where there was a significant heavy-server market in that
protocol.  NeXT doesn't sell high-end server CPU configurations,
and so there really isn't anything to hook up a NeXT Terminal to.

--
Scott Byer                             NeXTMail: byer@mv.us.adobe.com
Adobe Systems Incorporated               These are *my* opinions, and  
1585 Charleston Road, P.O. Box 7900        do not necessarily reflect
Mountain View, CA 94039-7900             the opinions of my employer.
---------------------------------------------------------------------

In article <SCOTT.93Jan22132602@nic.gac.edu> scott@nic.gac.edu (Scott Hess) writes:
>In article <1jp3mmINNnpo@cs.widener.edu>
>	jefferis@manas.engr.widener.edu (Raymond P Jefferis) writes:
>>In article <1993Jan21.160050.4997@tse.uucp>
>>	markt@tse.uucp (Mark Turner) asks:
>>> Is it possible to hang X terminals off a NeXT server and use the native  
>>> NeXT desktop from the terminals?
>>> Comments .... ?

This may be trite, but "A CPU on Every Desk", when taken literally, is a given,
whether the architecture distributes the graphics protocol (X), the
database processing (Client-Server), or the application processing
(distributed 3D rendering).  The question revolves around the design of the
OS and its services.

Having a fully-configured workstation at every user location has always
seemed extremely wasteful of resources, both in up-front costs and in support
and maintenance. The dumber the device, the easier it is to maintain.

From the perspective of managing data, distributed data models are beginning
to cause serious anxiety in corporations because much data is no longer
secure or readily archived.

Spreading out applications on end user work stations leads to
administration/licensing problems and more difficulty in supporting the
tasks of the user.

In general, these are problems faced by entities with large growing
networks and the poor blokes who have bought into PC networks as a
corporate strategy will be the first to feel the bite.

The issue goes beyond the distribution of the display processing. Data
storage requirements on a network are growing at very high rates so that
disk arrays can outgrow the capabilities of the OS.

The major tasks of the OS, (persistant data management, application
processing, display processing) could all be split into separate subsystems
that communicate across a "network".  This network could be standard
ethernet for some tasks, specialized buses such as Sun's Mbus, or high
speed DMA-like buses for data access as in Plan 9.

Robust protocols are required.  Each subsystem can run the core OS,
with specialized extensions for its primary task while communicating with
the other subsystems in a standard way such as UNIX has used from the
beginning with device files.

The issue is thorny and real.  X terminals are very attractive from one
point of view, but the drawbacks are just as compelling.  Standardizing on
X itself may be right for an organization to preserve software investments
since X will run on just about any CPU configuration that is found at a
user's work station.

I'd like to see a powerful, networked NeXTstep windowing system.  It would
allow many sites to buy basic NeXT hardware and put it to good use.  These
machines could even be diskless as long as they shipped with enough RAM.

Since this is the advocacy newsgroup, then put one vote in the yes column
for NeXT to plan for this in future releases.
-- 

Mark T. Dornfeld                              Voice: (416) 234-9048
CYANTIC Systems                               Facsimile: (416) 234-0477
101 Subway Crescent Suite 2103                Email: mark@cyantic.com

In article <1993Jan22.184704.20171@adobe.com>, byer@adobe.com (Scott Byer) writes:
|> Mark Turner writes
|> 
|> > I've come to the conclusion that a "CPU on every desk" is not the
|> > right solution for many businesses. As soon as a $20,000 workstation
|> > lands on a corporate desk, it needs more memory and more hard disk
|> > ... and the CPU itself it out-of-date and too slow in just a few
|> > months.
|> 
|> > The X terminals can be replaced, or upgraded, without much fuss
|> > and little expense.  And it's simply more efficient to do system
|> > administration, and manage most of the hardware problems, centrally.
|> 
|> Your question is making the implicit assumption that the above is
|> true.  Having watched X terminals in action, and seeing what they
|> do to a network, I don't believe that it's a valid assumption.
|> 
|> X terminals aren't any cheaper than a well-equipped 486, which can
|> then run Unix and X native, with lots of disk, and not swap over
|> the network.  A full Unix machine can also be easier to administer
|> in a complex, shifting environment.  (Not easy, mind you, just
|> easier than a X terminal.) Especially if using NetInfo.

the two previous paragraphs are interesting claims.  i disagree (surprise)
as do our customers and apparently much of the marketplace. a major
reason people buy X terminals is because they're relatively trivial to
administer, and getting easier all the time.

|> Also, CPUs are a cheap resource compared to networks.  This will
|> remain true for at least 5-10 years.  And X terminals simply _swamp_
|> the network.

this has often been asserted, but no one has yet backed it up with any
data.  several companies (including NCD, DEC and HP) have done studies
that show that X terminals have little effect on network traffic, even
with 'network killers' like Frame.

|> It just isn't effective to be running more than a
|> few on a single subnet.  And for responsiveness, you need an adequate
|> amount of peak computing power.  That ends up being just about one
|> CPU per terminal in the end.  That's in addition to the high-power
|> processor in the X terminal.

again, i disagree.  here at NCD we have >100 people all sharing a single
Solbourne.  i've watched tests with over 60 terminals pounding on a
single Solbourne with no loss in response.  we have customers doing
similar things.

[ and our low-end mono terminals use the 'high powered' 68000.  not all
our customers need the latest CPU. ]

|> --
|> Scott Byer                             NeXTMail: byer@mv.us.adobe.com
|> Adobe Systems Incorporated               These are *my* opinions, and  
|> 1585 Charleston Road, P.O. Box 7900        do not necessarily reflect
|> Mountain View, CA 94039-7900             the opinions of my employer.
|> ---------------------------------------------------------------------

speaking for myself,
-- 
Dave Lemke				lemke@ncd.com
Network Computing Devices, Inc.