[9fans] Sun Ray: Deja vu

[9fans] Sun Ray: Deja vu

[steve.kilbane]

Sun's web site is currently full of "Sun Ray", its devastating new breakthrough
in systems design, moving all the computation back into a central powerful
server, and putting dumb, exchangable bitmap/keyboard/mouse units on the
desktop.

Right.

It has a number of interesting aspects, not least of which is the bandwidth
required. Instead of using something like bit(3) to do rendering on the
terminal, Sun Ray has X clients render into a virtual frame buffer on the
server, which is then zapped across to the terminal. This might be one of the
reasons why Sun Ray requires a dedicated switched network, rather than being on
the LAN.

I didn't spot any mention of the distinction between cpu and file servers (since
the server runs Solaris, I guess they're combined). As far as this goes, there
is a solid border between cpu server and terminal: the terminal does no
application processing.

Authentication can be via normal login, or smart cards, with a smart card reader
built into the terminal base unit.

An interesting side-effect of using a proxy X server on the server is that
sessions need not be terminated when you detach. They can continue, and be
recovered when you log in again, at another terminal. Another devastating
breakthrough, although there is no mention of "Teleporting", which Olivetti Labs
did in the early 90s using proxy X servers.

Supposedly, Citrix WinFrame stuff can be used to display Wintel apps on the same
screen, seemingly at the same time, and with cut'n'paste between the Solaris and
Wintel apps. Plus, Sun are pushing StarOffice as a means of getting access to
most of what Wintel's needed for, without requiring Wintel client licences.

Roughly speaking, this seems to take the hardware/administration aspects of the
Plan 9 model, and apply them in a blunt manner to the existing Solaris system,
with no other changes.

I vaguely recall Brazil having a different graphics model, more suited to
exploiting high-bandwidth networks. That wouldn't be another unrecognised
ancestor of this, would it?

steve

[9fans] Sun Ray: Deja vu

[kim]

The following came across:
-----Original Message-----
From: steve.kilbane@ind.alstom.com <steve.kilbane@ind.alstom.com>
To: 9fans@cse.psu.edu <9fans@cse.psu.edu>
Date: Tuesday, October 05, 1999 10:55 AM
Subject: [9fans] Sun Ray: Deja vu


>
>Sun's web site is currently full of "Sun Ray", its devastating new
breakthrough
>in systems design, moving all the computation back into a central
powerful
>server, and putting dumb, exchangable bitmap/keyboard/mouse units on
the
>desktop.
>
>Roughly speaking, this seems to take the hardware/administration
aspects of the
>Plan 9 model, and apply them in a blunt manner to the existing Solaris
system,
>with no other changes.
>
>I vaguely recall Brazil having a different graphics model, more suited
to
>exploiting high-bandwidth networks. That wouldn't be another
unrecognised
>ancestor of this, would it?
>
>steve
>
and so:
(DANGER: Long Boring Message Ahead)
This idea of local processing for the interactive user
stuff and 'down the wire' processing of the big stuff
and the storage is nearly thirty years old. A lot
had the beginnings at Cambrdige (UK), but without the
benefit of the DARPA billions available in the US,
much of this work never got beyond the prototype/proof
of concept stage.

Obviously, there won't an exact one-to-one mapping for
systems using 1999 hardware (e.g. SunRay), but the
fundamental approach is quite similar.

For example, 'RAINBOW - A Multi-Purpose CAD System'
(Software-Practice & Experience (1972) v2, p239)
describes a distributed system with a PDP-7 on the
user desk (whopping 8K of store) with light pen,
stylus tablet, teletype, and CRT, connected to a
TITAN computing system  with 128K store, paper and mag
tapes, plotters/printers, etc. To quote, "give users
many of the facilities of fully interactive computing
without the overhead of stagnant programs in core
awaiting terminal I/O."

If you look at dmr's 'History of Unix' pages you'll
see that sperm-meets-egg at the scanned PDF of Martin
Richards BCPL manual, given to Ken Thompson (I assume)
while Richards was at MIT on sabbatical from Cambridge.
Richards was a rather unprolific publisher, but you
find his ideas pervading almost everything with a cpu
these days. I assume he lectured to Bjarne Stroustrup
and gave him a few pointers on techniques for handling
complexity in large systems, maybe even pointing him to
the 'Lauer-Needham Duality Hypothesis'.

At Xerox PARC during this period (the mid-1970s) Warren
Teitelman brought some ideas from MIT, e.g. 'A Display
Oriented Programmer's Assistant' (Int. J. Man-Machine
Studies (1979) v11, p157), an InterLisp system which
the user had a minicomputer to handle the bit-map display,
windows and mouse, with PDP-10/Tenex networked in to run
the Lisp interpreter. The Acme/Wily folks might recognize
the meaning of a quote from that paper: "any text that is
displayed on the screen can be pointed at and treated as
input, exactly as though it were typed."

And if you check AT&T's 'new' Cambrige Labs
(www.cam-orl.co.uk/ring.html) you'll see that "the first
high speed digital network platform was the Cambridge
Fast Ring." One project that utilized this slotted ring
was an update of the RAINBOW, this one using a 68000 cpu
and 2901 bit-slice processor to handle the display and
window system. This distributed system used Martin
Richards TRIPOS (Software-Practice & Experience (1979),
v9, p513) portable OS on the desktop machine, the file
servers, and the minicomputers in the 'cpu pool' or
'processor bank'. This gave invariant response time
for user interaction but will full scale processing
available down-the-wire'.

They even developed a fonts & typesetting system ('use ELF
for your elf-abets') and the Rainbow allowed "an
undergraduate project to enable eight different published
methods for performing the anti-aliasing calculation to
be compared. Only a windowing display with grey-level
capability could do this at all, and only the Rainbow
could do it as responsively." ('Computer Bulletin' (Dec.
1995, p. 25).

And to finish up, at PARC (they did a lot of cross-breeding
with Cambridge) Warren Teitelman published "A Tour Through
Cedar" (IEEE Software,(1984) v1, n2, p.44) which describes
a networked environment using the Dorado on the desktop and
'tiling' on the screen: "it is one of the most widely discussed
aspects of the Cedar user interface, and often leads to heated,
religious debates between its adherents and advocates of
overlapping windows."

And then Teitelman went to Sun Microsystems. And now we
have the fabulous 'new' SunRay. The future's so bright I
gotta wear shades.