copyright notice
accesses since April 8, 1996

Windows, Unix, OS/2 and the Mosaic War


[figure 1a] [figure 1b] [figure 1c] [figure 2] [figure 3]


Many of us are old enough to remember the excitement of the microcomputer revolution. There was a widespread feeling that something important was happening and that our civilization would be forever changed (for the good, we hoped). While an important technological milestone, the microcomputer revolution may pale in comparison to the revolution taking place in cyberspace.

At it's core, the microcomputer revolution was an important, but predictable extension of the "smaller, faster, cheaper" philosophy embraced by computer industry as a whole at the time. As it turned out there was a threshold in the level of circuit density which had to be passed in order for CPU-on-a-chip technology to be possible. Once that threshold was passed in the early 1970's, talented scientists and engineers, together with a few imaginative entrepreneurs, pushed the technology onto the desktop. Within a half-dozen years the software developers turned toys into office tools, and the rest, as they say, is history.

The current revolution in cyberspace is more unique in many ways. Cyberspace is a dimensionless electronic workspace - a place where knowledge and information sources come together via the ubiquitous digital common carriers. It is a virtual conduit for digital media of all types. In many ways it is as close a sibling to science fiction as high technology. (In fact, the term "cyberspace" came from William Gibson's science fiction novel Neuromancer). It wasn't as if we had miniaturized cyberspace, or extended it in some important respect. Nothing like cyberspace had ever existed before the mid-1980's.

There are three technological advances which allowed cyberspace to develop. First, there was the presence of a unifying digital infrastructure - in short, the networks. These digital common carriers are unifying because all transmissions, no matter what information they contain, are ultimately reducible to strings of binary digits or bits. The digital networks are the cement that holds cyberspace together. While there are many competing networks, the one which is the most important at this moment in time is the Internet.

The second requirement was the software support of an appropriate set of operational metaphors. Paramount among them is cut-copy- paste paradigm from desktop publishing. The support of the extensibility of such metaphors to non-textual domains then made multimedia a reality. One now routinely cuts, copies and pastes sound bites, video clips, animations and virtual slices of reality just as one manipulated words and phrases in ASCII documents just a few decades ago.

The third requirement was the software required to support virtual terminals. This concept is derived from the original remote login procedures built into Telnet many years ago. Now, GUI interfaces, multimedia and network navigation allow such programs with strange names like Mosaic, Harmony, Gopher and Prospero to escort us on virtual trips through cyberspace. What you see on your display terminal may only exist on your display, at least in that form. This is the world of client-server (aka distributed) computing.

Of all of the network protocols for such virtual travel, none has caught on like the World Wide Web. It represents the most rapidly growing sphere within cyberspace. The tools of the Web are the client-server navigators and browsers. Since we're representing the desktop perspective, we'll confine ourselves to the client-side of the cyber-fence.


Cyberspace has its origins in the Unix community. A respectable number of the major innovations in the client-server arena have appeared first on Unix platforms. This owes as much to the fact that many, if not most, of the high priests of cyberspace were baptized in Unix and remain faithful to the cause.

However, most of the world's future cybernauts will remain outside the order. The numbers speak for themselves. There are somewhere around 50 million licensed users of Windows 3.x (not counting the millions of pirates) and another five or six million OS/2 users who run Windows programs. Compare that to the few million Mac users and even fewer Unix users. The big-ticket cyberspace markets on the client side are obviously Windows markets.

While the client side of the Internet was originally dominated by Unix workstations, this is no longer the case. The PC/workstation dichotomy has all but vanished with the modern Intel microprocessors and multitasking PC operating systems (OS/2 and Windows NT), and the utility of cyberspace access has been convincingly demonstrated to large numbers of typical computer users (read that PC users).

The domination of the client side of cyberspace by PC users took place within the last few years. Current estimates peg Internet users at 20 million - several times the number of Unix workstations in use. It is projected that this figure will double within the next 12-18 months. Most of the projected growth of Internet use will have to come from the 125 million or so PC users, since the Unix world is already connected for the most part. We're talking mega-market here for both Windows and its client-server software. (Actually, there is even a larger, though shrinking, market for DOS client-server technology!)

The point is that the operating system which prevails in desktop computing will inevitably prevail over the client side of cyberspace. The developers will orient their best work for the largest markets and hence the largest potential revenues; in a word, Windows. It won't really matter that NT has a 1980's look- and-feel, or that Chicago 95 may not ship 'til 96. Windows' installed base will drive the market not because of any stellar technology but because of close-knit relationships with third- party developers and an extremely aggressive approach to preload contracts with OEMs.

So that's where things stand in January of 1995. Cybernautics is a schizophrenic world of both Unix and Windows apps.


Enter OS/2. Left out in the cold? Hardly. From the point of view of cyberspace, 1995 could become the golden age for OS/2, although it isn't at all obvious how long that might last. OS/2's strengths are three-fold.

First, IBM produces a solid native OS/2 X-server client which integrates seamlessly with Unix client-server environments (although the initial setup is a hassle best left to the pros). There is also OS/2 support for Motif. In short, OS/2 has X-apps pretty much covered for any Unix affectionado who doesn't want to wander too far from the well. They can become one with OS/2 without much discomfort.

Second, most of the major client navigator/browser development activity at this moment is in the Windows 3.1 market. WIN-OS/2 supports most of these 16-bit net apps except for those that are tailored for Windows NFS, OLE2, etc. - currently few in number). To the degree that Warp remains compliant with Windows standards, especially NFS, OLE2 and Winsock32, it may remain competitive for 32-bit client apps as well. In other words, for the moment OS/2 has the Windows 3.1 market covered which, for all intents and purposes, is most of the Web navigator/browser market.

Third, OS/2 is a superb multitasking environment for the desktop. Client side software, especially when it comes to cybermedia, consumes a lot of system resources. With a navigator/browser loaded, a multimedia player spawned, and some serious file exchange over the net via NFS mounts, a computer without preemptive multitasking soon becomes a boat anchor. One of life's little ironies is that the big market for multi-threaded clients currently for Windows 3.1. Multi-threading for Windows 3.1 - what is wrong with this picture? Writing a multi-threaded app for Windows 3.1 is like trying to tailor a dwarf - there just isn't enough to work with. To be fair, the developers are actually postitioning themselves for the Windows95 and NT market rather than Windows 3.1, but you couldn't tell it from their promotional material. None of the developers as near as we can tell seem to be taking the MacIntosh market as much more than an afterthought.


For these three reasons, excellent X-server support, respectable integration with Windows 3.1 clients, and a robust multi-tasking environment, OS/2 has a lot to offer today's cybernauts. In our view OS/2 is the only way to go at the moment. Will it remain that way? We have doubts.

First, IBM has no serious native OS/2 presence in the client-side market. IBM's Web Explorer follows IBM's long-standing tradition of being late to market with stale technology. Web Explorer is to cyberspace what the RT was to workstations. It would have received top honors in the fourth quarter of 1993. Regrettably, it didn't ship until the fourth quarter of 1994. Therein lies the rub.

Figure 1a, 1b, 1c. Web Explorer, Air Mosaic and InternetWorks concurrently launched from OS/2 Presentation Manager

That's not to say that Web Explorer isn't any good. It's a sturdily built, useful client that just came out too late to make any difference. The backplane seems to be rock-solid and the interface is tasteful and mainstream. The problem is that the product was out of date before it shipped. In order to see why, we need to look closely at the cyber-surfing experience.

Cyberspace is an interlinked collection of information resources of all types. A typical surfing experience might involve concurrently playing an MPEG movie from a distant server in one window, using another window to monitor file transfers over the net, previewing digital artwork in a third, downloading newly discovered shareware in a fourth, cutting and pasting document fragments to the clipboard in a fifth, building a print queue in a sixth, and all of this while other desktop apps are printing and spooling and formatting and so forth. There is a lot going on.

Figure 2.Multithreading and Multipaning on InternetWorks running under OS/2
Taking advantage of the wealth of cybermedia available in real time places heavy demands on both system and personnel. And in order to surf efficiently one needs a high-performance client. We will discuss four performance boosters available in current high-performance Web navigator/browser clients.

First, multithreading in a Web client is a must. This is the ability to simultaneously load and view several documents. This is important for two reasons: the bandwidth over the Internet produces long load times (especially during periods of peak usage) and a high percentage of links lead down blind alleys. So if cybernauts wants to avoid spending much of their time watching documents load, they need multithreading.

Second, there is so much information in cyberspace that one really needs to automate the search process. Integrated search engines which look for key words and phrases in resources are starting to appear in the newer clients. This can quickly locate interesting resources without actually requiring a manual perusal - an absolute must given the size of cyberspace.

Figure 3. Search Capability Integrated into Netscape Client
Third, the standard way of retaining lists of interesting resources is with hotlists or bookmarks which contain the addresses (technically, uniform resource locators or URL's). These lists of clickable entries are the surfer's personal rolodex for cyberspace. Without them each access requires typing out a string of uniform resource locator information (the familiar "http://....." business) each time a resources is accessed.

The problem with hotlists and bookmarks is that they don't scale well. The first fifty or so entries, sorted alphabetically, don't cause much trouble, but beyond that it's a problem. For long lists, it may take as long to find the URL as it takes to load it. For that reason, most advanced clients now offer folder management for hotlists. The better clients also offer editors and support URL annotation. Time-savers, all.

Finally, the reality is that it is the graphics and not the text of documents that takes up the better part of the load time. Modern clients are moving toward deferred image loading which loads the text and links first so that the user can continue the navigation through cyberspace, and the graphics last. This feature offers the cybernaut some protection from bandwidth bandits (documents laden with graphics) lurking around the nets.

These are four imporant performance features which give modern clients a competitive advantage when they compete for the Windows desktop. The following table compares Web Explorer with some of the leading Windows products in terms of these features:

                                   (1)  (2)  (3)  (4)  (5)

     multithreading                 -    -    +    +    -
     integrated search              +    -    -    -    -
     hotlist/bookmark mgr.          +    +    +    +    -
     deferred graphics loading      +    +    +    +    -


(1) = Netscape (v1.0; Netscape Communications Corp.)
(2) = Air Mosaic (v3.0; Spry Corporation)
(3) = Internetworks (beta 4; Booklink Technologies)
(4) = Win Tapestry (v1.67 ; Frontier Technologies)
(5) = Web Explorer (v.91; IBM Corp.)

This table says it all.

Of course, Web Explorer also excels in some things. Reconfigurable caches for both documents and graphics is an interesting innovation, as is image rendering. The kiosk or full screen mode is very useful for presentations because it eliminates the distractions introduced by the background desktop. But while these are nice features, they are unlikely to be perceived as mission-critical by future cybernauts. The the performance features are likely to be far more important to the informed cybernaut. In this regard, Web Explorer just isn't competitive.


Where just a few years ago, the World Wide Web and the National Center for Supercomputing Applications' (NCSA) client, Mosaic, were inextricably linked, today they are moving apart at breakneck speed. NCSA is currently licensing the use of the tradename and code to some third-party developers, but many of the better innovations are coming from independent, non-licensed developers. Of the products compared above, only Spry's Air Mosaic is an NCSA Mosaic descendent.

This departure from the Mosaic standard suggests a healthy degree of anarchy amongst developers. This is surfacing in more innovative software and better integration with existing mainstream desktop apps. This bodes well for the cybernaut as these navigator/browsers become increasingly convenient to use. In the last year or so almost all of the significant advances on the client side have come from the Windows developers, with Unix a distant second, and MacIntosh an afterthought.


IBM is not likely to become competitive in the client navigator/browser market. As a standalone program, Web Explorer appears to be a cut above average when compared to the entire spectrum of Unix, Windows and MacIntosh clients. However, it is not competitive with the better Window's programs as a standalone Web client, and looks even worse when it comes to integration with NFS, email, news reader and conventional desktop apps like word processing, desktop publishing and multimedia.

In our view, OS/2's future in cyberspace is dependent upon IBM's ability to ensure Windows compliance in subsequent releases, particularly when Windows95 ships. At this writing there are only two platforms which can take full advantage of the more advanced features of the better Windows clients: OS/2 and NT. In this arena, OS/2 remains very competitive. If OS/2 maintains Windows compliance, it will not be hurt by the rapid advances in cyberspace products for Windows - it may not be helped much, but it won't be hurt. In other words, OS/2's strength in cyberspace will be linked to Windows products just has it has been in officeware. We predict that Web Explorer, like Top View, will linger awhile and then gracefully fade into Cyberspace lore as a product that was too little, too late.


(* denotes seamless integration with OS/2 2.1; + requires Winsock32 compliance which will not run under OS/2 2.1 but which may run under Warp). Tested on IBM ValuePoint 486DX2 with OS/2 2.11, IBM's TCP/IP suite and generic 8-bit ethernet card.)

[author's bio]
Hal Berghel is an active researcher in many areas of experimental computing including the design and engineering of client-side networking software. He is also an active freelance writer on cyberspace and cybermedia topics, including a regular column, "Cybernautica", which appears in PC AI. His URL is