Issue 1-40, September 11, 1996

Be Engineering Insights: The Devil in the Board

By Pierre Raynaud-Richard

[This is a fiction based on real events. Any similarity to actual people or machines will be... perhaps intentional.]

Friday, 9:23 pm

An anonymous Be geek, alone with his computers, is reading

The BeBox is a great machine... except for graphics card support which is really miserable.

"Why is Be supporting only the 964 (which is no longer available), and not the 968 (which is). I *know* they're almost identical..."

"Is Be so lazy that they can't even support the Matrox Millenium?"

Dagnabbit, these guys are right. What was Be thinking when they decided to do such a bad job with graphics card support? I need to help them... At least they had the intelligence to release the sources for the few graphic drivers that they do provide. Let's see... here I have a Diamond card based on a Vision S3 968, using VRAM. A pretty good card. Too bad they don't support it. I never did a graphics driver before, but it's probably no more difficult than any other piece of software.

I'll plug the card into my BeBox and turn the machine on. Uh oh, the screen is black. No Be logo, nothing. That's strange. When I plugged this card into my Wintel machine, everything just worked—I didn't even have to copy a new driver onto my disk. Why can't I do the same thing here? Wait... Of course! God, I'm stupid! The Wintel uses an 80x86 processor which can execute the BIOS code of the ROM. The BeBox would need an emulator to do the same thing, and that's not such an easy job.

But it's not a big deal. I have the source from Be for a 964-based board. I just need the databook of the 968... Here it is.

[Be geeks are extremely resourceful.]

Great, they explain the differences between the 968 and the 964—just what I need. This should be an easy job.

Friday, 10:28 pm

Tarnation! Still nothing! All the settings look good, but the screen is still completely black. What could be wrong? I'm using the TVP 3026 RAMDAC/clock generator; the Be driver uses the TVP 3025. That's probably close enough—but maybe I should check to see if they're really compatible. I should have those databooks somewhere... Here. OK, let's compare everything...

Friday, 10:51 pm

Fuor la spada! These guys are completely crazy! They rename or move half of the registers, and half of the others were completely modified. No wonder nothing worked. Probably I can try to patch the code used to program the TVP 3025...

Saturday, 12:04 am

Okay, now I'm making progress. There are still huge bugs in the graphic settings, but at least the screen isn't black anymore, and the monitor was able to synchronize correctly.

Saturday, 3:11 am

O Entsetzen! These are nasty bugs...

Saturday, 5:43 am

It's too late now, I'm getting nowhere. Time to sleep.

Saturday, 1:32 pm

Let's see if those wretched bugs can evade me now that I'm fully awake!

Saturday, 8:58 pm

Okay. Maybe I'm not going about this the right way. These chips probably have bugs that aren't documented. I need more information. I should be able to get driver sources from somewhere else. Let's try Linux and Windows NT... Here they are! First Linux... hum, the source is 1 MB... and a bit confusing. What are they doing ? Oh... it looks like they're still using the old bank-switching mode. What about Windows NT? Oops! The source is even bigger (1.3 MB) and it's even more confusing. Huge macros everywhere... links to their global graphic system... and more source code.

Linux, Windows NT... in both cases it looks like it will take days or weeks to understand what they're doing and reproduce it. And they're not even supporting some of the advanced features used by the Be drivers. Probably the best is just to look for specific comments about the 968.

Saturday, 9:17 pm

I'm going nowhere again. Too big, too complex, too different. My driver is probably almost working, except for a few bugs. Stupid to begin everything again from scratch... Okay, let's try disassembling the Windows driver.

Sunday, 6:02 am

Looks like another big failure. They're doing strange things in a strange order. And the API is so complex! I hope I never have to write a driver for that OS. Off to sleep. Maybe I'll have another idea tomorrow...

Sunday, 2:16 pm

I was doing things the wrong way. What I need to do is just put the card on my Wintel machine in a mode that's as close as possible to the one used on my BeBox and then dump the values of the registers. Then all I have to do is check what my BeOS™ driver is doing. Pretty simple. I should have done that in the beginning.

Sunday, 3:57 pm

Criminidly! I've set the registers to be as identical as possible, given the differences in the settings between the machines. But it's still not working! Svenami! There are probably bugs in the program order too! I remember that there's a case in which you have to first set register A to 0 before you set register B to a specific value, and only *then* could you set A to its real value... Dumping the registers won't help here.

But I have one more solution: Where's my logic analyzer?

[We told you the geek was resourceful.]

I'll plug the analyzer into my Wintel machine and spy on the transactions between the CPU and the graphics card. I can only store one set of 1024 samples, and I need 2 samples per transaction. But 500 transactions should be more than enough. Lets see what it's doing when it turns on the graphics card in default mode during the boot...

Sunday, 4:04 pm

Uh oh, it looks like there are going to be more than 500 transactions. Why so many? Strange. Bah, there are probably just a few more. Let's keep tracking...

Monday, 12:19 am

I can't believe it! I can't believe it! I've already looked at 40,000 transactions, and there are still more... Just to boot the card! What's going on? Is there a devil in the board? What's it doing? Can a virus have reprogrammed the BIOS ROM? My God! The guys who wrote these pazzo drivers are completely crazy! Wait... I know one of them! Perhaps I can get a few hints out of him. It's not too late to call...

"Hi John, it's Paul. How are you?... Me too, me too... No, I called because I was just trying to adapt a driver for your 968-based graphics card... Yes, that one... How long!?... It took two engineers an entire year to get the Wintel driver working?!... I understand better now... A lot of bugs? I can believe it... So thanks John, see you soon."

Perhaps I chose the wrong card. Better to try another one. Oh, here's my Matrox Millenium. Where's the databook? How could I not have the databook?

[Remember, the geek always has everything he needs.]

Oh, I remember, I tried to get it two months ago, but John said even they couldn't get it. Matrox just doesn't give it out. Pretty annoying... Looks like I'll have to wait for Be or a third party to give me new drivers. They said they're working on it...

Be Developer Talk: StarCode, Hard at Work

By Carlin Wiegner, Michael Klingbeil

Like many other Be developers, StarCode Software jumped at the chance to develop for this new and exciting platform. The BeOS represents an intriguing mix of old and new. The Be interface is a strange blend of Windows, Macintosh, and even some UNIX interface ideas which seem to work well together. The stability and performance of preemptive multitasking and memory protection are a welcome change on the personal computer frontier. The small, flexible Be API is much less intimidating than the numerous "Inside Macintosh" volumes. The database and multimedia support are full of tremendous potential.

As many people were quick to point out, Be was crazy to try to build another OS when the world seems to embrace only a small handful of OSes. What Be had going for them, whether they knew it or not, was an OS that would appeal to a wide range of users, from geeks and hobbyists to multimedia designers. And with the introduction of the PowerMac port, Be has an opportunity to reach the mainstream.

Even though there's a lot of work left to be done on the OS and certain aspects of development are still frustrating, we should start seeing a change on the applications front over the next six months. Be is in a perfect position to aggressively track down great products from other platforms. Some areas where we see wonderful opportunities for small shops are: A hybrid resource/database editor, some memory/heap tracking tools, a news reader, and a graphical FTP client.

Located in Menlo Park, California, StarCode Software specializes in writing software management tools for the Be platform. We will be releasing the first version of our software installer, the Package Builder, during the second half of September. This is a critical step for developers to have a better tool than tar to distribute their software. Preliminary versions of the Package Builder will be free while we close in on a commercial quality product. (Feel free to suggest features and report bugs.)

Following the Package Builder, we will begin developing a second product, the Software Valet, which will help users spend less time managing their software and more time using it. Early versions will keep track of a user's installed applications (and concomitant software "packages"), thus allowing the user to easily and cleanly uninstall particular programs (for example). Later versions will track software compatibility, offer software upgrade assistance, and automate registration.

We're excited to bring these tools to the Be platform. For more information, send us e-mail at or check out our web site at By the way, we're looking for another engineer. If you're interested, see the job announcement on our web site.

Be Marketing Mutterings

By Alex Osadzinski

I well recall, in a reminiscence that will date me as a startup ancient, my painful technolust for the portable Z80- based CP/M machine called the Osborne 1. I saved up my L, s, and d in anticipation of purchasing my own Osborne 1—until Adam Osborne, the company's eponymous founder, announced the second-generation machine somewhat ahead of its shipment date, with the unfortunate effect that sales of the first product ground to a halt as everyone (myself included) waited for the newer, far superior, Osborne 2. The company failed and bequeathed its legacy to the computer industry:

Os-borne (Awz'born) vi. to cause a decrease in or elimination of sales of a product by announcing its replacement prematurely.

Despite this, companies continue to announce—or leak—information about new products before they are available. Sometimes this is simple misadventure, but there are other motives for early announcement, some pure, some not so pure.

I'm prompted to write about Premature Announcement Syndrome because we may have unintentionally been unclear about the nature of two recent announcements: Our licensing of OpenGL® from Silicon Graphics and our agreement with Metrowerks to develop Java tools for the BeOS. Our intent was to inform about our future direction.

In both cases, our announcements met two important criteria: They were news and they were factual. We had indeed concluded a license agreement with Silicon Graphics and had indeed agreed with Metrowerks that they would develop Java tools for the BeOS. Neither was a product announcement, but both have been interpreted as such by some people. We apologize for the confusion. Our intent was, and still is, to indicate two important directions for the BeOS: That it will support the OpenGL® API, making the porting and development of OpenGL®-based applications much easier; and that we are planning to support Java, an important emerging language.

We'll announce both products when they ship, with a target for both of them late this year or early next year.

Samuel Adams, Again...

By Jean-Louis Gassée

There are a number of reasons why I like Samuel Adams, or Boston Beer, the brewery. It makes excellent beers and, for Be, an excellent metaphor —a little outside my usual range perhaps, but an example of a well-run company that succeeds in a market dominated by giants. Recent events have rekindled my interest in the metaphor and also provided an answer to observers who kindly teased me by pointing to a flaw in the imagery.

The first event is a very nice story written by Steve Kaufman, a staff writer at the San Jose "Mercury News." The piece appeared September 9 and, thanks to the net, was brought to my attention by friends over in France. Titled "Companies Putting the 'Public' Back into IPOs," the column describes how companies such as Boston Beer or Starbuck Coffee make sure small individual investors do not get shut out of "public" offerings. Boston Beer allocated 20 percent of the 4.5-million-share offering to customers who called for a prospectus and the opportunity to buy shares at $15 each. What's remarkable isn't that shares still trade above the IPO price (quote symbol SAM: $21 today), it's that Jim Koch, Boston Beer's CEO, took the extra pain and money required to include consumers in the IPO. It took an extra shot of pain in dealing with the SEC, getting their approval, writing a special consumers-only prospectus, and spending over a million dollars in the process. Apparently, this was money wisely invested. The consumer/shareholders turned out to be Boston Beer's best promoters, they got cards identifying them as Boston Beer co-owners and everywhere they went, demanded to know why the store or the restaurant they were in wasn't carrying Samuel Adams beer.

This isn't the way most companies do IPOs. In order to get stock in a hot IPO, you need friends in high places. Either at the company or in one of the firms handling the new issue. The little guy stands outside that cozy network and only gets to buy after the insiders and their cronies made their profits on the initial rise. Too bad if the stock falls back not only from the early apex, but also below the offering price, as it did often in recent hyped-up offerings.

This is not a new phenomenon. In 1980, I got a call from Aaron Orlhansky, a financial analyst from a US brokerage house located in Paris. We used to meet for lunch, he brought press releases and prospectuses, I interpreted the technical jargon, and he picked up the tab. One day, by way of thanking me for pointing Tandem to him, he offered to get me stock in the upcoming IPO of Apple Computer. I declined, preferring to stick with my small Tandem position. He also mentioned that Tom Lawrence, then Apple's VP of European Operations, was looking for someone to start Apple France. This time, I bit.

Back to Boston Beer, I used to like the company because it made an excellent case for both contrarian wisdom and plain old good execution. Back in the seventies, the Boston Consulting Group and their brethren sold concepts such as "forward pricing" and "size is everything". As a result, we witnessed ugly copulations of large corporate bodies, megaconcentrations in the food, tobacco, and beverage businesses. Anheuser Busch bought competitors, Philip Morris bought Miller breweries, and we got corporate beer, beer designed by committee, indifferent beer. Less taste, more filling. On the consumers' taste buds, this left room for more interesting beers. Microbreweries, an old pioneer tradition, started to rise, so to speak. We got mediocre beer, so-so beer, and great beer. Boston Beer rose from nowhere, initially with little support from the financial community, with no manufacturing plant. But their beer consistently won taste contests. They contracted capacity from other breweries, stayed in control of the taste, the quality, and the cost, and initially marketed the hell out of their product at beer festivals and other connoisseur venues. Samuel Adams and the other fine microbrews became so successful the giants woke up irritated and confused. Their response? Imitation. By carefully reading the label of one such ersatz, you'll find it's made under a license from a French industrial brewery. The same attention to a Plank Road label will reveal it comes from the makers of Miller Lite.

I was introduced to Boston Beer by one of our founders, Steve Sakoman, home brewer himself, who felt Samuel Adams was one of the few commercial brews worthy of real beer makers and drinkers. So, after metabolizing a few pints, I started metaphorizing. Boston Beer and Jim Koch became one of the examples that inspired our work. That was long before they went public. Not only did Boston Beer execute well against conventional wisdom, but when the time came for a hot IPO, they also treated their consumer/shareholders warmly and creatively. Reading Steve Kaufman's column made me feel we had indeed picked a good model.

What about the flaw in the imagery, then? Astute observers remarked this microbrewery metaphor bucking contrarian thought was well and good, but all beers are compatible with the installed base of refrigerators and bottle openers. Our beer needed not just taste buds tired of legacy brews, but also different refrigerators and bottle openers. It just occurred to me that with the PowerMac port of the BeOS we might have found nice standard fridges and church keys.

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:


Subject: Sound card discussion (or why my BeBox sounds better than my PC)

AKA: Be Audio Stats

More discussion of sound cards vs in-box audio. This week we had comparisons with Gravis Ultrasound and Soundblaster AWE-32. In a related thread, a call was made for AES/EBU connectors on the BeBox. Also, there was some speculation on Be's GM synthesizer.

THE BE LINE: We will be releasing the General MIDI synthesizer through our Web site in a few weeks. Stay tuned.


Subject: Single and double clicking

AKA: Clicks, doubles n' drags
AKA: Multiple Clicking
(and so on)

A number of mouse issues:

  • When the user double-clicks the mouse, what messages should be sent to the target application: Two single clicks? A single click followed by a double-click? One double-click? Should the application be able to specify which of these it wants?

  • Should a double-click always do something "constructive" (such as open a document), or can it be destructive—can it be defined to quit an application, for example?

  • What's a "click": Is it a mouse-down, or a mouse- down/mouse-up?

  • How should a three-button mouse be emulated by a mouse with only two buttons? Should the user be allowed to choose the form of emulation?

  • When a dragged item is dropped, how should the source (window) be informed of the item's new location?

Subject: RunFilePanel and RunSavePanel

It was suggested that the open and save panels allow broader file filtering and format conversion. Currently, the BApplication class lets you filter files based on their types (only).

Subject: Floating point performance

The clarion has sounded: Use floating point. It's fast, it's easy, and it's (sometimes) free. Not much controversy here—but some cautions: Division is slow, as is int<=>fp conversion. Pierre Raynaud-Richard offered a handy rundown on the comparative speeds of various operations in various contexts.

Subject: Bug Errata for BeOS would be wonderful.

Many developers wrote in to cast their vote for publicly disclosed bug reports (from Be). Many of these developers have been asking for this sort of support for quite awhile. Speculation is that Be doesn't have the necessary technology...

THE BE LINE: The hang-up hasn't been technology as much as manpower stretched thin. But you're right—you deserve to see bug reports and we will give them to you soon (honest). Look for an announcement within a couple of weeks.

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