NeWS

From Wikipedia, the free encyclopedia

NeWS, for Network extensible Window System, was a windowing system developed by Sun Microsystems in the late 1980s. Its primary architect was James Gosling, who subsequently designed Java.

Based on PostScript (PS), NeWS started by modifying the PostScript interpreter to run in a cooperative multitasking fashion. Unlike PostScript in a printer, NeWS would be displaying a number of PS programs at the same time on one screen, so some form of multitasking was required. Despite their similarity in using PostScript, the NeWS interpreter was unrelated to Display PostScript.

NeWS added a complete view hierarchy system, based on viewports known as canvases. Like the view system in most GUIs, it included the concept of a tree of embedded views along which events were passed. NeWS also included a complete model for these events, including timers and other "automatic" events, input queues for devices such as mice and keyboards, and other functionality required for full interaction.

But by far the most interesting addition was a complete object oriented (OO) programming style with inheritance. This eliminated the need for an external OO language to build a complete application.

Since all of these additions were extensions to PostScript, it was possible to write simple PostScript code that would result in a running, onscreen, interactive program. For instance one of the common examples was an onscreen clock, which required about two pages of code.

NeWS also included a library (several actually) of user interface elements (widgets), written in NeWS. These widgets ran all of their behaviour in the NeWS interpreter, and only required communications to an outside program (or more NeWS code) when the widget demanded it.

For example, a toggle button's display routine can query the button's state (pressed or not) and change its display accordingly. The button's PostScript code can also react to mouse clicks by changing its state from pressed to not pressed and vice versa. All this can happen in the windowing server without interaction with the client program, and only when the mouse is released on the button will an event be sent off for handling.

This was more sophisticated than a "dumb" X Window System server, which can only report "mouse was pushed down here", "mouse is now here", "mouse was released here" events to a client, which then has to figure out if the event is in the button, switch the state, and finally instruct the server to display the new state. If client and server are not on the same machine, these interactions must travel over the network, slowing the feedback loop down unnecessarily.

The best example of such a toolkit is TNT (The NeWS Toolkit) which was released by Sun in 1989. Sun also shipped a smaller toolkit intended for example purposes and making small programs.

Several companies licensed NeWS and adapted it for various uses, but usage certainly couldn't be described as widespread. SGI used it to replace their proprietary GL windowing system. The OpenLook version of FrameMaker developed by Frame Technology Corp. with funding mainly from Sun Microsystems and NSA was one of the few commercial products successfully run on NeWS.

The freely-available X11 was already quite popular, so the first versions of NeWS emulated X11 by translating the calls into NeWS PostScript. Speed problems plus the existence of programs that relied on the exact pixel results of X11 calls, forced Sun to release an X11+NeWS hybrid called Xnews which ran an X server in parallel with the interpreter, which seriously degraded the NeWS interpreter performance and did not really result in a very good X server either. Sun also attempted to emulate the look & feel of OpenLook with a toolkit built on the same Xt (X Intrinsics) base as Motif, called OLIT, so that X programs could be made to look identical. After it was clear that OpenLook had lost out to Motif in popularity, and after Adobe acquired FrameMaker, products on NeWS simply vanished. Most UNIX workstations (including Sun's) now run the X Window System.

[edit] Why did NeWS fail?

There is no doubt that in many ways NeWS had a superior design for thin-networked clients, by moving much of the processing to the display, and separating graphical user interface semantics from client program semantics.

Possible reasons for its failure in the market include:

NeWS needed to be licensed from Sun, while the source code for the X Window System was freely distributed under the MIT License. Any commercial code shipped using the NeWS libraries required licensing fees to be paid to Sun, Adobe Systems, and Xerox PARC.
NeWS lacked a robust library of reuseable code until well after the X Window System had become the dominant paradigm. This mistake was obviously not repeated in Java. Making matters worse, the variety of widget sets offered by Sun simply confused the users.
NeWS tried to make PostScript programming easier by adding object oriented extensions, but this just confused most software developers. Several "compilers" from C-like syntax such as pdb (PostScript Done Better) and c2ps were available, but were cumbersome to use and not supported by Sun.
Writing NeWS apps required coding the client- and server-side parts of the application in two very different programming languages. Coordinating the communication between the two sides was sometimes found to be difficult.
The implementation of the NeWS window server never achieved the level of robustness of competing window systems. The situation was made worse with the NeWS/X11 merge, which was bloated and caused the workstations of the time to swap.

It is interesting to contrast NeWS with Display PostScript (DPS), which used the same underlying imaging model and language, but did so in a very different way. In DPS the PostScript commands were limited to what was needed to draw things, all other operations (such as creating a window to draw into) had to be implemented using other system interfaces. In comparison with NeWS, DPS lost interesting features like the ability to use a PostScript path to describe the shape of a window, and also meant DPS required you to use the quite horrible (but familiar) Xlib and some truly ugly code to make sure both DPS and X were agreeing about what to do. However it also meant that the majority of the system and applications use it was located in compiled code, making it many times faster and considerably easier to write and debug. The result was a much smaller engine that offered the same end result — a PS-based display — but with higher performance and somewhat more "natural" programming.