Turns out round window with stroke path text alpha and reimplemented xeyes was not compelling.

People should have shown NeWS doing complex client side form field checks.

Sun also didn't go downmarket. Windows PCs happened with hummingbird xceed X not decent in Microsoft X

Having the toolkit in the server would necessarily force it to define and enforce lots of policy. This increases complexity, increases dependency (coupling) on part of the clients, and after all that you still have the problem that policy simply doesn't age well.

It still boggles my mind a bit that a display stack designed in the 1980s has remained viable to this day. Sure, it's a patchwork, and it's finally being replaced, but it works.

The big drawbacks of NeWS in my opinion are not having an MIT license, and being quite difficult to implement well. Also PostScript isn't the nicest language to program in, which I say from years of programming in it. But it's far from the worst, and if one changed the to postfix syntax to prefix, it would be mostly a Lisp.

There is nothing more multi-core than a modern GPU. Actually the existence of modern GPUs is an argument FOR toolkit-in-server windowing systems. All you do is writing shaders that help you to keep the communication over the PCIe bus (and potentially network, if you want network transparency) as minimal as possible.

The NeWS system failed for none of the reasons described in the article. It failed because of bad performance. There never was custom silicon out there to render Postscript. Unlike today where you have vastly underused GPUs at your disposable just waiting to be stuffed with intelligence in form of shaders. Just have a look at https://shadertoy.com to see what is possible even on low-end GPUs.

Edit: as cpach pointed out shadertoy.com and not .org

Among other advantages, I can imagine a world where PostScript would be used for building user interfaces: a document in screen and print would look the same, without need for writing conversions like, e.g. HTML+SVG -> PDF when someone needs to somehow print a web form quickly. And perhaps the old principles of typography and the experience in typesetting beautiful books with PostScript could be more easily applied to UI design.

I searched Wikipedia and found that "Great White Hope is a nickname or epithet originally applied to boxers who were the hope of the "White Race" to beat African American boxer Jack Johnson."

That seems an odd way to characterize the situation.

NeWS was architecturally similar to what is now called AJAX, except that NeWS coherently:

- used PostScript code instead of JavaScript for programming.

- used PostScript graphics instead of DHTML and CSS for rendering.

- used PostScript data instead of XML and JSON for data representation.

FTFY:

Web Browsers were a mistake, as are all toolkit-in-browser JavaScript libraries

April 1, 2020

One of the great white hopes of the part of the Web world that never liked Gopher was Netscape's web browser. Never mind all of its practical flaws, all sorts of people held web browsers up as the better way and the bright future that could have been if only things had been different (by which they mean if people had made the 'right' choice instead of settling for Gopher). One of the reasons people often give for liking web browsers is that it put much of the JavaScript libraries into the web browser instead of forcing every server to implement it separately.

Unfortunately for all of these people, history has fairly conclusively shown that web browser was a mistake. Specifically the core design of putting as much intelligence as possible into the web browser instead of the web server has turned out to be a terrible idea. There are at least two big reasons for this.

The first is parallelization. In the increasingly multi-core world you desperately want as much concurrent processing as possible and it's much easier to run several web servers in parallel than it is to parallelize a single web browser. Even if you do get equal parallelization, separate web servers are inherently more resilient because the operating system intrinsically imposes a strong separation of address space and so on, something that's very hard to get in server where everything is jumbled together.

(I believe that this is one reason that modern HTML CSS rendering has been moved from the web browser to the web server. WebFont rendering is increasingly complex and CPU-consuming, so it's better to stick web servers with that burden than dump all of it on the web browser. After all, all Web Servers should have multiple GPUs for drawing and accelerating graphics and numeric computing for all their clients, but there is no rational reason anyone would ever want to program the GPU in the web browser, or push that processing off from the honking big accelerated server farms that you're renting from Amazon by the minute, to the millions of web browsers run by your users whose electricity you don't need to pay for. Plus, users enjoy waiting for their graphics to download over the internet in response to every mouse click and movement they make. It gives them time to think and reflect.)

The second is that if you put the toolkit in the web browser you make evolving the toolkit and its API much more complicated and problematic. The drawback of having everyone use the web browser toolkit is that everyone has to use the same web browser toolkit. Well, not completely. You can introduce a mechanism to have multiple toolkit versions and APIs all in the same web browser and allow web servers to select which one they want or need and so on and so forth. The mess of a situation with the current web browser and its JavaScript libraries make a very educational example of what happens if you go down this path; not very much of it is good.

(Some JavaScript libraries are in practice mandatory but still must be probed for and negotiated by the web servers, while others are basically historical relics but they still can't be dropped because some web server somewhere may ask for them.)

JavaScript libraries in the web browser push the burden of dealing with the evolution of the library into the web server. It is the CDN and web browser cache that carries around old or new versions of the library, with various different APIs, and you naturally have old library versions (and even old toolkits) go away entirely when they are no longer used by any active web servers (or even any installed web servers, when things get far enough).

(I'm ignoring potential security issues for complex reasons, but they may be a good third reason to be unhappy with browser-side libraries.)

(This was part of a discussion with Owen about NeFS, which was a proposal for the next version of NFS to run a PostScript interpreter in the kernel. More about that here:)

https://news.ycombinator.com/item?id=17077721

Owen Densmore's discussion of John Warnock's "Linguistic Motherboard" idea for PostScript:

https://donhopkins.com/home/archive/NeWS/linguistic-motherboard-owen.txt

    Date: Tue, 20 Feb 90 15:20:52 PST
    From: owen@Sun.COM (Owen Densmore)
    To: don@cs.UMD.EDU
    Subject: Re:  NeFS

    > They changed the meaning of some of the standard PostScript operators,
    > like read and write, which I don't think was a good idea, for several
    > reasons... They should have used different names, or at least made ..

    Agreed.  And I DO see reasons for the old operators.  They could
    be optimized as a local cache for NeFS to use in its own calcs.

    > Basically, NeFS is a *particular* application of an abstraction of
    > NeWS. The abstract idea is that of having a server with a dynamically
    > extensible interpreter as an interface to whatever library or resource
    > you want to make available over the network (let's call it a generic
    > Ne* server).

    Very true.  This has been particularly difficult for me to get across
    to others here at Sun.  I recently wrote it up for Steve MacKay and
    include it at the end of the message.

    > It's not clear to me if NeFS supports multiple light weight PostScript
    > processes like NeWS.

    I asked Brent about this, and he agreed that it's an issue.  Brent
    has been talking to a guy here who's interested in re-writing the
    NeWS interpreter to be much easier to program and debug.  I'd love
    to see them come up with a NeWS Core that could be used as a generic
    NetWare core.

    I think you should send  your comments off to nfs3 & see what happens!
    I agree with most of your points.

    Owen

    Here's the memo I consed up for MacKay:

Window System? ..NeWS ain' no stinkin' Window System!

-or-

Swiss Army NeWS: A Programmable Network Facility

======================================================================

Introduction

NeWS is difficult to understand simply because it is not just a window system. It is a "Swiss Army Knife" containing several components, some of which contribute to its use as a window system, others which provide the networking facilities for implementing the client-server model, all embedded in a programmable substrate allowing extremely flexible and creative combination of these elements.

During the initial implementation phase of the Macintosh LaserWriter software, I temporarily transfered from Apple to Adobe working closely with John Warnock and other Adobe engineers. At lunch one day, I asked: "John, what do you plan to do after LaserWriter?" His answer was interesting:

        PostScript is a linguistic "mother board", which has "slots"
        for several "cards".  The first card we (Adobe) built was a
        graphics card.  We're considering other cards.  In particular,
        we've thought about other network services, such as a file
        server card.

This paper explores NeWS in this light: as a Programmable Network Facility, a major part of Sun's future networking strategy.

NeWS Networking Components

NeWS has realized John's notion of a programmable network facility. It has populated the initial PostScript mother board with several cards vital to netwoking:

    -Device Independent Graphics Primitives: The key importance of the
     PostScript graphics model for networking is that it is device
     independent, able to represent with one file the same drawing on
     *all* the network hosts.  The Folio fonts contribute to this: they
     can be rendered at a given point size on all screens, regardless
     of resolution.

    -Host Independent Interpreter: PostScript is an interpreted language
     that runs on all Sun workstations without recompilation.  The
     significance of this for networking is that small code fragments
     can run on any host without any knowledge of the host's
     architecture.  Thus one host sending a code fragment to another
     host need not be concerned whether that host is a Sun-3, Sun-4, or
     even Macintosh!  This extends to client programming languages as
     well: all client languages have the same interface to the host.
     As an illustration of this point, both C and Lisp clients use
     the TNT toolkit with no special interfaces for each language.

    -Network Socket Primitives: NeWS has added primitives for listening
     for connection requests, for socket files, for host name and
     internet number management, and for reading and writing on the
     network connection.  In fact, the NeWS "server" code is not
     written in C, it is written in PostScript.  The initial version,
     and the one I run on my own machine today, is 10 lines of code!

    -Light Weight Processes (LWP): A network host has many independent
     activities taking place at one time.  Each of these uses a
     separate LWP.  These activities need not be window oriented,
     and in fact most are not.  Consider WUE alarms, for example.
     A LWP could be created on each WUE NetStation whose only task
     was to listen for activities on other WUE NetStations and to
     cause a WUE Alarm to occur when an interesting event occurs.

    -Garbage Collection: When data is allocated within NeWS, it is
     returned to the free memory pool when no process references it.
     This is needed in a multiprocessing network host because separate
     LWPs can reference the same shared data.  (No process can safely
     free data itself, because other processes might reference it.)
     Garbage collection manages this problem for the processes, and
     thus is an important Network facility.

    -Programmable Events: NeWS has a flexible, very easy to use
     event management scheme.  Processes fill out templates of events
     they are interested in.  When these are matched by events occuring
     within the system, they are delivered to the interested process.
     Events are *far* more useful than simple keyboard and mouse events.
     They can be artificially manufactured for any sort of inter-LWP
     communication.  For the previous WUE Alarm, for example, the WUE
     "Listening" process would communicate to WUE-Friendly applications
     by filling out a WUEAlarmEvent and sending it.  The interested
     applications would have filled out a template which would "catch"
     the event.

    -Programmable Error Handling: NeWS uses the standard PostScript
     exception handling, extended to work within the context of LWPs.
     Thus an error in one LWP is isolated to that process only.
     Network programs can "wrap" incomming code inside of the special
     "stopped" error handler which will catch errors occuring inside
     the code fragment.  This is vital to safe network programming.

"Well, I'm glad NeWS has all these useful parts, but just what IS NeWS anyway; how do I get at and use all this stuff."

Well you may ask! It does seem confusing.

Formally, NeWS is simply a Unix command, just like "ls" and "cd". Its syntax is:

    xnews [PostScript-code]

As a silly example of handing in your own commands to NeWS, we can do this:

        bigmac% xnews "1 1 add ="
        2

Now that we've gotten though the fundamentals, the rest of the paper will look at individual examples of using NeWS in novel ways. Although we present programs to illustrate how things are done, the reader may skip over them, the accompaning text will explain what the program does.

The Ten-Line Server

As mentioned above, "xnews" is simply a Unix command that executes a PostScript program, defaulting to "(NeWS/init.ps) run". Init.ps performs several initialization tasks and executes the commands in various other files. When its done with all of that, it defines and executes a small procedure called "server". This program uses the NeWS networking facilities to build a LWP which listens for connection requests from "client" programs. Whenever a request is heard, the LWP awakens and creates and initializes a new LWP just for the client's usage.

Here's the ten line version of the server. This actually works (I've run it for quite some time!), and is explained in Chapter 10, Networking NeWS, of the Unix Networking book published last year.

    /server { % - => -; create the NeWS client-server listner process
        {   clear newprocessgroup           % -;    Clean stack, init process group
            (%socketl2000) (r) file         % soc;  Open socket for listening
            {   dup acceptconnection        % soc f;Loop forever creating clients
                {   exch pop 200 dict begin % f;    Create userdict
                    initmatrix newprocessgroup % f; Init graphics & process
                    cvx exec                % -;    Execute the connection!
                    currentprocess killprocessgroup%Cleanup client & children
                } fork pop pop              % soc;  Clear process & file
            } loop                          % soc;  Loop forever creating clients
        } fork pop                          % -;    Fork server & toss process
    } def

ps2bits: A PostScript to Rasterfile Program

A second use of running xnews as a Unix filter is illustrated by the ps2bits probram. This Bourne shell script reads a PostScript file on its "standard in", and emits the resulting Sun raster file on its "standard out". The shell script does this by having xnews:

-Read in a few standard PostScript initialization files,

-Initialize the PostScript environment (& fix a Folio bug!)

-Build an 8.5 by 11, color, 72 dpi resolution "page" canvas

-Execute the PostScript stdin file to draw that page

-Write the resulting image to stdout as a raster file

Here is the program:

    #! /bin/sh
    xnews "
        (NeWS/pack.ps)      (r) file cvx exec
        (NeWS/basics.ps)    (r) file cvx exec
        (NeWS/redbook.ps)   (r) file cvx exec
        500 dict begin false setautobind () stringwidth pop pop
        612 792 8 [1 0 0 -1 0 792] null buildimage setcanvas
        (%stdin) run clippath (%stdout) (w) file writecanvas
    "

Once the image is made, it can easily be looked at in SunView, XView, or NeWS using any of their raster file viewers. Here's a trivial viewer, "bitwin", using the TNT toolkit:

    #! /bin/sh
    psh <<BITS
        /win [ClassCanvas] [] framebuffer /new OpenLookBaseFrame send def
        {   clippath pathbbox scale pop pop
            (${1}) readcanvas imagecanvas
        } /setpaintproc /client win send send
        (${1}) /setlabel win send
        100 100 425 550 /reshape win send
        /activate win send
        /map win send
    BITS

To use these on the demo file "tiger.ps" which was initially created on the Mac using Adobe Illustrator, these two steps are taken:

    bigmac% ps2bits < $XNEWSHOME/demo/PostScript/tiger.ps > /tmp/bits
    bigmac% bitwin /tmp/bits &

Although the complexity has been fairly minor in the previous examples, we've illustrated some powerful capabilities:

    -We can easily use NeWS to provide a simple Unix facility for
     converting PostScript drawings into Sun raster files.  The
     PostScript may come from any of the many widely available sources,
     even the Macintosh.

    -These raster files can be used by any toolkit, including SunView.
     They can even be used by other raster devices such as printers.

    -Using the networking primitives used in the 10-line server, we
     could easily build a similar server for providing raster files
     to anyone on the network.

WUE Alarms with Programmable Events

WUE has the notion of "alarms" which can be used to alert applications of changes in status of network objects. Applications can use this for anything from "hot links" to file status alarms. One of the WUE examples is on file status: I want to know when the monthy report reaches 90% completion. Hot links are similar: I want this graph to change whenever the associated spreadsheet changes.

Here is one way to implement WUE file alarms using NeWS programmable events:

    -WUE provides a Property Sheet which includes a file completion
     field holding a percentage quantity.

    -When the Property Sheet is completed, it creates and sends a
     WUEStatusEvent for the given file.  The network name of the object
     is stored in the event's Action field.  The event includes
     auxillary information about the file in the ClientData of the
     event.

    -Applications following WUE object status changes use the NeWS
     "expressinterest" primitive to catch all WUEStatusEvent's for the
     desired file.  It then looks at the ClientData to determine
     whether or not further action should be taken.

    createevent dup begin
        /Name   /WUEStatusEvent def
        /Action (/home/project/status/january) def
        /ClientData 1 dict dup begin
            /Completion .70 def
        end def
    end sendevent

    createevent dup begin
        /Name 1 dict dup begin
            /WUEStatusEvent { % event => -; check if complete yet
                /ClientData get /Completion get .90 ge {
                    AlertApplication
                } if
            } def
        end def
        /Action (/home/project/status/january) def
    end MyEventMgr expressinterest

Puny WOF (Workgroup Object Facility)

The above alerting works only for applications running on the same workstation. This is because the event that is sent is only visible to interests expressed on the same workstation. We can easily extend our model to include sending the event to many network workstations by using the same techniques as the 10-line server:

    -WUE friendly workstations start a WUE event listening service when
     they start up NeWS.  This uses exactly the same techniques as the
     10-line server, but using a different socket file.  They also
     regester themselves with a centralized list of WUE friendly
     workstations.

    -The WUE event listening service starts a second process which
     expresses interest in all local WUE events.  When any WUE
     event is caught, it is "forwarded" to each of the workstations
     listed in the central list.  It does this by opening a socket
     file to each workstation, and sending a small program down
     the socket which sends the WUE event on the remote machine.

    -The Property Sheet sends the WUEStatusEvent exactly as before.
     Local applications running on the same machine operate exactly as
     before.  In addition, however, the local WUE event forwarder
     causes the same event to be sent on each of the regestered
     workstations.  Applications on those machines would now catch the
     forwarded event exactly as if it were local.  The application need
     not be modifed!

Multi-Media: NeWS Device Control

Multi-media devices can be managed by NeWS, using programmable events to interface to applications. This provides a logical division between the device "driver", which may be implemented in any of several ways, and the application API to the device, which is via NeWS events.

As an example, consider a graphics tablet input device with an RS232 interface. This device emits x,y data in a well-known format to the serial device when the tablet's stylus moves. As an initial implementation, the driver half of the device was written in PostScript as a process which opens and reads the file /dev/tty. When there is no data on the device, the process blocks, pausing until data arives for the device. When the data arives, the read returns with the device data. This is packaged into an event as in the WUE Alarm example.

Typically the device data needs processing that is better suited to C programming. The solution is to rewrite the device manipulation as a NeWS client using either CPS or the TNT Wire Service. The application interface, however, continues to be the same event interface. The improved performance incurrs no change in the application interface. The notion of this variety of client "daemon" has been used in TNT for providing shared text service among applicaions. A voice interface to EMacs was also made in AD using the same technique.

Another significant advantage to this division of labor into a device layer using an event interface to applications is that it gracefully evolves into sharing the single device among multiple applications. The device is temporarily "owned" by one application, and is given to another application using an agreed upon protocol. Two common styles are used:

    -The keyboard uses a "focus" model where certain UI events not
     associated with the keyboard determine which application owns it.

    -The mouse is typically owned by the application in which it was
     last "clicked".

Although device ownership may seem like a trivial problem, fairly complex systems are typically needed to solve it for a given device. The VOX server from Olivetti for voice is used for just this kind of thing: negociation between applications for ownership of the voice device. Because VOX is not integrated into a more general event system, however, it cannot manage the user interface for which application currently owns it. It can only provide the regestry. When OpenLook defines the "focus" model for ownership of the voice device, a NeWS oriented voice service will much more easily integrate into the rest of the system.

Just as WUE Alarms had a network component, certain multi-media devices will be expensive enough to warrent sharing on the network. Scanners and fax machines will be attached to individual machnes, but we may want them to be temporarily owned by other workstations. The same "server" technology used by the 10-line server and the Puny WOF can also be used here.

Summary and Vision

The interesting software architectures of the next decade will leverage network technologies. We will see WUE become the Net Finder, replacing the simplistic Mac Finder in the network intensive workstation world. WUE will manage network resources among WUE-Friendly NetStations, workstations with advanced, easily configured network software platforms. NeWS is such a network tool: a programmable network facility. Another is the newly proposed NFS which uses file server extended PostScript as its client-server language.

We hear far too many NeWS vs. X arguments based on window and UI toolkit arguments alone. These, in my opinion, miss the main strengths of NeWS. It is a programmable network facility which can be used for many purposes. It can prototype new network services. It can be a universal device controler. It can be used to integrate multiple hosts. All these elements also need to be evaluated when considering NeWS.

References

[1] PostScript Language Reference Manual, Adobe Systems, Addison Wesley, 1985

[2] The NeWS Book, Gosling & Rosenthal, Springer-Verlag, 1989

[3] Unix Networking-ch 10: Networking NeWS, Owen Densmore, Hayden Books, 1989

[4] NFS-3 Design Document, Brent Calaghan, Sun internal document, 1990

Appendix A: A more general ps2bits shell script.

    #! /bin/sh
    # Parse cmd line args: ps2bits [-size W H] [-dpi X Y] [-color]
    X=72; Y=72; W=8.5; H=11; Z=1
    while [ ${#} -gt 0 ]; do
        case "${1}" in
            -dpi)   shift; X=${1}; shift; Y=${1} ;;
            -size)  shift; W=${1}; shift; H=${1} ;;
            -color) Z=8 ;;
        esac; shift
    done
    # Run NeWS as Unix filter. Note use of Bourne shell variables!
    xnews "
        (NeWS/pack.ps)      (r) file cvx exec
        (NeWS/basics.ps)    (r) file cvx exec
        (NeWS/redbook.ps)   (r) file cvx exec
        500 dict begin false setautobind () stringwidth pop pop
        $X $W mul $Y $H mul $Z [$X 72 div 0 0 $Y -72 div 0 $H] null buildimage
        setcanvas
        (%stdin) run clippath (%stdout) (w) file writecanvas
    "

(Good work getting traction on this kick! ;-)

I'm kind of dealing with something like this now myself. I have an implementation in Prolog of a Forth-like language called Joy and I want to make a GUI for it that might actually be used by normal people.

I could use e.g. SDL and draw everything myself and handle mouse events etc. (I have a crude demo in PyGame that does that.)

Or I could use some sort of widget toolkit (I've got a Tkinter demo too.)

Or I can use the browser and treat it like a smart display processor. There are editors and spreadsheets and libs and all kinds of stuff. The Prolog/Joy code can stay nice and high-level, with HTML and CSS as the "display language".

- - - -

Which brings me to my main point: How in the heck is this not more nailed down in 2020!?

I first wrote a DSL for GUI layout in the late 90's, I used Delphi was even earlier, and I known that's all late to the party: we could go back to The Mother of All Demos or even earlier.

The mouse and hi-res raster displays and GUIs are old but usage is nowhere near settled yet. Is that because the domain is really so intricate and challenging or are we just milling around and mistaking it for progress?

Whither React?

(By which I mean maybe functional reactive programming or better yet Kowalski's Logic Production Systems (LPS http://lps.doc.ic.ac.uk/ ) might finally start to reduce the churn and nail things down.)

( https://en.wikipedia.org/wiki/Delphi_%28IDE%29 )