Issue 1-48, November 6, 1996

Be Engineering Insights: The Evil Empire?

By Erich Ringewald

Much has been written, and even more said, about those dark forces from the Northwest, the evil willies of Redmond. Why does one little software company evoke such harsh sentiment from almost everyone in the industry (not to mention the occasional investigation from the Department of Justice)?

The subject came up again at dinner last night, and I was unsurprised by the unanimity and depth of feeling. "Unscrupulous bastards," "criminals," "lunatics." Round the table we went (myself included), in a game of "top the epithet." I started to feel sorry for Bill Gates.

But... I've witnessed my share of Microsoft's "interesting behavior." Is it coincidence that Microsoft announced PenWindows just one day before Go Corporation announced its PenPoint operating system for portable pen computers? Microsoft was a key PenPoint developer—Go had even invited Microsoft to demonstrate their application at the PenPoint rollout. Yet, somehow, the other half of Microsoft (the system software folks) chose to repay this kindness by scooping Go with their own PenWindows intro. This "coincidence" took a little of the steam out of Go's announcement.

Criminal behavior? I wouldn't like it if they did it to my little sister, but... probably not illegal. Besides, wasn't Go asking for trouble by inviting the fox into the preproduct introduction henhouse?

I see Microsoft as a remarkably persistent company whose business tactics are extremely aggressive, bordering on ruthless. They push the limits, but as any adolescent can tell you: That's What Limits Are There For. Microsoft has a position of power in this industry that's unparalleled, and they sometimes abuse it. But it wasn't always this way...

[WAVY LINES ACROSS THE SCREEN; HARP ARPEGGIOS ON THE SOUNDTRACK]

I first worked with Microsoft when they had a single product, a BASIC interpreter. Not the stuff the world's largest personal fortune is made of, but it was a good product. It actually worked. Then they made their second product, a FORTRAN compiler for CP/M. It too was a good product. It also worked. And it was documented. Again, this may not sound like much, but it did serve to distinguish the product from its competition. Microsoft was a scrappy little company (about the size of Be), but it was just one of a number of scrappy little companies in the microcomputer language market. In the years before their big break with DOS for the IBM PC, Microsoft slowly but steadily increased their product line, their quality and support, and their revenues.

Microsoft first showed a version of Windows in 1983. It was miserable. Nobody bought it. They regrouped and refined and launched Windows 2.0. It too was miserable. Nobody bought it. They regrouped and refined and launched Windows 3.0 in 1990. Better. People started to actually use it. Refinements known as Windows 3.1 made it a highly successful product. Once again the traditions of persistence and refinement paid off.

There is a bumper sticker about Macintosh that really bugs me: "Windows '95 = Macintosh '89."

It's not the logic that bugs me—yes, of course Windows '95 is no better than Macintosh in 1989. But I want the companion bumper sticker, which is just as true: "Macintosh '95 = Macintosh '89."

So next time people start complaining about the tactics of Microsoft, I might be compelled to point out that while Bill Gates is our only modern example of the robber baron, his company's drive and persistence in continuously improving their products might also be part of their success.


News From The Front

By William Adams

What is a Technical Evangelist anyway? Well, it's a name that I made up so that I could fit into this awesome company. A Technical Evangelist is a person with a lot of engineering experience who has a fire in the belly to become a marketing type person. The evangelism aspect of the job is to spread the gospel of the BeOS™ as far and wide as possible.

The BeOS provides a platform upon which developers can "do the right thing." As a company that's interested in longevity, and ready market acceptance, we're interested in getting software from well-established companies, but that's not the exciting stuff. What's really exciting are the offerings of the small developers who take a risk on a new platform and dare to create something a little bit different from the past—and often a few levels better.

So, how can a Technical Evangelist help this process? Well... the Technical Evangelist can write code. A lot of code. The Evangelist can show developers how easy (or hard) it is to use the system in a particular way for maximum effect.

In working with NeXT and Taligent, I always pressed hard for good documentation and sample apps. So I believe this is the best thing for the BeOS as well.

This week finds us looking at how to support something like Photoshop. We have a modest application, Rraster! The primary function of this application is as a replacement for the ImageViewer application, which comes with the BeOS. I took a look at the ImageViewer code and thought, "This could be made into a lot better and more useful example."

Rraster! is like ImageViewer in that it lets you drag and drop images on it to view them. The twist is that it allows you to drag and drop images of any type! It achieves this by use of add-ons. The interface is easy, and a GIF and PCX add-on are included. There is a simple add-on manager that shows how add-ons can be easily supported in any application. I don't want to spoil all the fun, because this was also the basis for an upcoming magazine article, but there's a lot of good stuff in here.

The other things you'll see in this sample are how to properly launch an application and deal with parameters, whether you dragged and dropped onto the icon, double- clicked a file, or launched it from the command line. These are mundane issues, but they can be maddening if you can't do them right.

You can get the code from: ftp://ftp.be.com/pub/Samples/Rraster.tgz

The add-on support is very basic, and is meant to give you a good idea of how easy it is to use add-ons. As far as data translation in general is concerned, you should probably take a look at the DataTypes library by Jon Watte.

Tools For You

Writing graphics device drivers is a black art on a good day. If you have been trying to do this of late, you know what a pain it can be. Well, our graphics driver person has given me a tool that anyone writing graphics drivers will probably find useful. The tool allows you to test your driver like a normal app by hooking it up to an app_server dummy. You can do source-level debugging without having to reboot all the time, and when you're done, you can make it a full-blown driver. This is the very tool that we use ourselves, so you have it as good as we do. If you really want to get your hands on it, please send e-mail to me (wadams@be.com) and ask for it. I won't make a general public release, because the code is not very... pretty. But we'll give it to those who are really interested in exploring the heart of darkness.

Rallying The Troops

The BeOS is a fun and exciting operating system to write applications for. The proof is in the number of applications that show up on our ftp site every day. Developers at this stage are sharing and cooperating with each other. This is a good thing and will only serve to make applications better in the near term rather than later. We're currently creating a sample CD that will go out to all our demo teams, and thanks to everyone who sent in code to get on this disk!

More opportunities abound! Comdex is coming, Macworld is coming, and many more opportunities will give our developers exposure. Keep those apps coming! The more we can replace our own apps with yours, the better the BeOS demonstrates. We're in a much better position if we can show developer applications instead of our own. So keep them coming, we'll put you in front people's faces.


Be Developer Talk: Richard Deken

By Richard Deken

My primary occupation is as a VLSI design engineer. My responsibilies are a combination of circuit design and CAD software design and modification. I've been fascinated with electronics since I was a kid. As I dabbled in computer programming a few years later, I developed interests in scientific software tools developement. What I've enjoyed most is the continuous challenge and the constantly expanding base of knowledge from which I could learn exciting new ideas.

In the late 80s I closely watched the development of the Amiga computer system. It was a novel system with many features unmatched in other systems of the time. A group of friends and I decided that it would be cool to develop software for this system. We formed a small group that collaborated on a few small projects. Unfortunately the Amiga entered a long period of stagnation, and there was no attractive alternative platform for our varied projects.

Last year we started hearing rumors of the upcoming BeBox. At first we cautiously stood back and watched Be, Inc. The BeBox and BeOS seemed to be well thought out. Designed from the ground up, not just for multitasking, but for multiprocessing. The BeBox also had many nice standard features, such as SCSI and IDE, PCI and ISA, PowerPC CPUs, and the GeekPort™. (Clearly more than the cheapest common denominator sold by many other companies.) However, a good product alone is not sufficient.

We wanted to make sure that the company was going to be committed to developing a strong and unique product, and most important, that it was willing to commit to helping developers and to pushing the development of products on the new platform. Be seems to be doing precisely that. They seem to be willing to accept and support small-time developers. It's always been my experience that the majority of software used on systems is not the major applications, but the small additions, utilities, and tools developed by small developers. It's also these small things that give a system the general functionality (if not the feel) that makes a system so great. They also seem to be very open with information about their system. The major manuals and information about the BeBox and BeOS all seem to be freely accessible via the Internet.

I feel that developing for the BeBox is a unique oppurtunity. The user base is clearly a cut above average, for being willing and able to work with a new system such as this. Further, the market is probably going to be much more open to new software, users aren't likely to be limiting themselves to the "I've got the same software as everyone else" attitude. This will hopefully give not only software my friends and I develop a better chance of making it off the ground, but should also provide us with a more diverse set of software and tools for us to choose from to aid our work.

The BeBox also provides a lower-cost route to parallel processing. I look forward to investigating parallelizing software algorithms, and integrating these techniques in more responsive applications. The BeOS seems much friendlier to multithreaded applications than any other OS I've worked with.

The BeOS also seems to provide several other unique features that it would be cool to use. Some of these include the object-oriented approach used throughout the OS and the integrated file system and database.

Having only been accepted as a Be developer in the last few weeks, I'm still investigating my options as far what products I (and probably several of my friends) decide to work on. Numerous ideas I've been considering include:

Most likely none of these will be available before the middle of next year. Most of these are targeted at electrical engineers and hobbyists. They would probably be released as low-cost commercial packages.


Be Marketing Mutterings: Recruiting

By Alex Osadzinski

I like bananas. In fact, I like them a lot. I bring a few of these wonderful fruits to the office every day and my colleagues have probably noticed that I munch on one sometimes as I wander from cube to cube. I hope that they don't mind when I absentmindedly drop banana skins into their waste baskets. I guess I should retain my banana detritus and dispose of it in the Be kitchen, where it could aid in the evolution of the strange lifeform that is beginning to write code with pseudopods as it grows in the refrigerator.

Whether or not my Be colleagues have noted my passion for bananas, I know someone who definitely has: The management of a hotel in Hong Kong.

Last year, I had the pleasure of visiting Hong Kong eight times, to work at the Asian HQ of my then employer. On each occasion, having survived the truly unique final approach and landing to Hong Kong's tired, overloaded airport, I would present myself at the same hotel's checkin somewhat worse for wear and ready for bed. The first time I stayed there, I was greeted professionally and quickly checked in, and my room featured a little bowl of fruit. During the next 48 hours, I consumed, of course, the two available bananas but eschewed the kiwi fruit and apples. During the next seven stays, the level of greeting escalated from the regular checkin staff through to the manager herself and the bowl of fruit (and room) grew bigger and, on my last stay, included no less than eight bananas and a note indicating that more of them were available on request.

Imagine what kind of customer service system notes guests' banana eating patterns and adjusts the room inventory accordingly. And, although I have my peculiarities, they pretty much limited themselves in Hong Kong to unusually high banana consumption. Imagine, if you will, how this hotel's customer service system adjusts to a much wider range of guests' preferences and idiosyncracies. Impressive, and not a little scary.

Be's first distribution channel is direct sales of our products through the web and other methods involving direct contact between the customer and Be. When you order a BeBox, send in a check, ask for the shipping date, deal with warranty issues, or have general product and logistical questions, you're working with Customer Service. Right now, that usually means working with Lilia Simbe, who's built a reputation with our customers for quick, pleasant, useful responses. We're building our customer service function to enable us to deal personally and efficiently with our growing customer base while not losing that personal contact between the customer and someone who knows Be and its products and cares that customers are satisfied. Early next year, when we ship our BeOS for PowerMac product, we expect (and hope!) that our unit volume grows by two orders of magnitude, and we very much want to keep direct contact with as many customers as possible. That's why we'll continue to focus on direct sales while simultaneously building other channels for our products.

Direct customer contact is not, by the way, a way to make more margin on products. Life being what it is, and the "no free lunch" law continuing to apply, all channels tend towards generating the same bottom-line profit for a company. The main purpose of direct customer contact is to see, hear, and feel what customers think about Be and our products.

So if you get in touch with Be next year to order the BeOS for PowerMac and someone asks you if you like bananas, don't be too surprised.


Apple Rumors

By Jean-Louis Gassée

I receive many e-mail messages asking me to comment on rumored conversations between Apple and Be. No, no comment, maybe, yes—there are many possible responses. Rather than picking one, I'll direct you to two URLs:

http://biz.yahoo.com/finance/96/11/05/aapl_3.html
http://www.infoworld.com/cgi-bin/displayStory.pl?96115.eamel.htm


BeDevTalk Summary

BeDevTalk is an unmonitored discussion group in which technical information is shared by Be developers and interested parties. In this column, we summarize some of the active threads, listed by their subject lines as they appear, verbatim, in the mail.

To subscribe to BeDevTalk, visit the mailing list page on our web site: http://www.be.com/aboutbe/mailinglists.html.

NEW

Subject: Architecture Legacy

AKA: Are files really what you want?

The Be philosophy calls for a break from "legacy" software. But is Be true to the call? This long-winded thread was a thoughtful (if somewhat abstract) look at certain appropriations that, perhaps, shouldn't have been made. The centerpiece was the concept of files: Are files (containers of "sequential bytes of data") obsolete? Does Be pledge a misguided allegiance to "unstylized" ASCII?

Subject: Filesystem/database thoughts

AKA: Are files really what you want?

In anticipation of the DR9 filesystem rewrite, many developers are offering proposals of their own (and hoping out loud that Be is listening). Too many ideas here to summarize—many of the messages were quite detailed—but the gist of most of the offerings hinged on whether the filesystem should be "natively" hierarchical or "databased."

In a related thread, a contributor posted the results of the IOZone benchmark (which measures disk I/O speed). The results were not glorious, but judgement is being withheld until the new filesystem appears.

Subject: Database thoughts (not filesystem related)

Some general questions about the database lead to observations and theories of query performance as the database swells with records. The highest number of records reported was 60,000 (with no appreciable performance loss). Good, as far as it goes, but some folks think billions of records (2^31) would be a reasonable and practical test.

Subject: Exponential Rolls out X704, etc etc

Some thoughts on the recent announcement of chips that run at N zillion MHz. Will they fry the user? What will the cost be after you've added in the requisite memory? Does the average user even need that much speed?

Subject: caches & CPUs

This thread was spawned from the "Why aren't two threads better..." affair that started a couple of weeks ago. Here, we concentrated on how multiple CPUs share the contents of the individual caches, whether one big cache is better than many small ones (yes), and, somewhat unrelatedly, whether a thread can (or should) pass a pointer into its stack to another thread (not a great idea).

Subject: GUI

This highly speculative thread discussed the possibility/advantage of a 3D user interface. Would it be fast enough? Would the user intuitively understand how to fly around 3D objects? Would it just be a toy? (The answer to this last question: Yes... but what GUI isn't?)

Subject: Injecting input into the app_server stream

William Adams (the Be Evangelist) solicited specs for event data that developers would like to "inject" into (and so be able to detect from) the app_server. Developers responded and also debated different approaches to various event detection techniques.

Subject: Fresh Programs

Should programmers concentrate on porting existing programs, or, instead, take the ideas of the old and rewrite them to take advantage of Be's "fresh" software? Many folks wrote in to point out that you have to start somewhere—porting is important if you want to get a set of basic utilities up and running quickly. The thread then veered into scripting.

Subject: Fresh Programs

Subject: A software License Manager Library

Some words about protecting your product from piracy, the desirability of source code escrow, and the pains of hardware-specific software. This lead to some difference of opionion regarding the merits of free software. Does "free" mean "not good enough to pay for?"

Subject: 603e vs DSP

AKA: DSP Performance Comparison
AKA: DSP on the BeBox

The performance of PPC chips was compared to that of dedicated DSP chips. While it was admitted (in theory) that DSPs are faster than the PPC, it was generally agreed that programming a DSP is a pain. Also, one contributor found that the PPC is actually faster than most (15 out of 19) DSP chips (see http://www.bdti.com/articles/bench.html).

Subject: Sockets

The thread started with a simple question about select(): Are you limited to 32 sockets because of the 32-bit fd_set mask? (No.) This led to a discussion of whether Be should abandon the BSD socket API in favor of an object-oriented interface (or similar high-level layer). Opinions were divided: Some folks find the BSD calls clunky and limited, others think that these calls are here to stay—why accept the overhead of a cosmetic layer?

THE BE LINE: The BSD approach was taken to make it easier to port existing socket code. A higher-level interface is being consider (time permitting).

Creative Commons License
Legal Notice
This work is licensed under a Creative Commons Attribution-Non commercial-No Derivative Works 3.0 License.