Barlow and Lavin


Lavin: NeXTSTEP is not just an environment to work in, like the Macintosh OS. It's a house in which you live.

NeXTSTEP 1.0 was a medieval castle. It was glorious, huge, and beautiful. But it was a little drafty, lacked a few amenities, and had a true torture chamber in the dungeon. Version 2.0 was a mansion in the city. While there were still concerns, they were the concerns of the wealthy. Not, 'Will we have a place to sleep?' but, ‘Is the duck pond full?' There were lots more amenities and most of the obvious flaws were repaired, but some anachronisms remained.

Now we're getting ready to move into 3.0. NeXT is assuring us that by the time you read this, 3.0 will be shipping. My guess is that the contractor misestimated by a bit, and your move-in date has slipped. But so goes homebuilding and software production. You'll be home soon.

Wait a minute. NeXT hasn't moved us into a new house. NeXTSTEP 3.0 is really a home improvement project. NeXT has added a few wings onto our old home and done some minor redecorating throughout – but your life will change completely only if your work takes you into the new wings. If you don't swim, the new pool is no big deal.

I think, though, that all of us will have occasion to deal with a few of the new features. AppleTalk and Novell support is a boon for everyone who works in mixed environments, as is the ability to seamlessly insert Mac disks. DBKit should finally allow some diversity of database applications, and I look for cool things from 3DKit.

The redecorating job in the main house is tasteful. The new color environment is beautiful, the uniform help system has been needed for some time, and the links between applications should affect us all in the future. A lot of minor bugs are gone as well. A few things still sound creaky, but overall the remodeling job will make users happy. At least while they dream about that flashy 4.0 house.

Barlow: There's an old advertising maxim that says, "Sell the sizzle, not the steak." Since there are few sizzle salesmen more adept than Steve Jobs, many of the jaundiced old hands in the audience at NeXTWORLD Expo were watching with both eyes when he described, at great length, the general shape and feature set of NeXTSTEP 3.0. Most of us came away convinced that 3.0 would be as stirring as its predecessors.

It turns out: Not! As I write this, I've had a few days to play around with a fairly precarious beta of 3.0 and all I can say is: "Where's the sizzle?" While there's definitely plenty of beef here, 3.0 generally has all the sex appeal of dangling carcasses.

Well, perhaps that's not quite fair. It does have one characteristic often associated with arousal: good looks. If you have a color NeXT, you're going to find yourself working in the most aesthetically pleasing environment ever to grace a video display. The newly colorized icons are lush and without a trace of tackiness. This interface is buffed.

Nor will anybody call the rest of 3.0's features and benefits superficial. Indeed, many of them are so far beneath the surface of things that many users will never know they're even there: Things like PostScript Level 2 and Pantone color matching, NeXTlinks, network faxing, e-mail encryption, software compression, broader printer support . . . . I mean, this stuff is basic.

The painful fact is that 3.0 leaves out a lot of the things I'd hoped to see in it – like fixes for the many minor flaws of NeXTmail, or an efficient method of searching for lost files on a big disk, or the ability to become root on the graphical level without logging out, or a better way of handling large numbers of fonts, or better inter-nal support for macro generation. My list was long and unfortunately it remains so.

Nevertheless, 3.0 does seem like a very well considered foundation to support both NeXT's short- and long-term strategies. For the short-run service of the custom-apps market, it contains improvements like DBKit, enhancements to Interface Builder, distributed objects, Novell and AppleTalk support, and on-line help.

More exciting are enhancements like 3DKit and software JPEG compression. These seem to be aimed at a much faster CPU than the 68040. So I guess if I'm looking for sizzle from 3.0, I'll have to wait for a processor hot enough to get all this meat cooking. Thumbs up, sort of.

Barlow: As I keep insisting, I'm not a UNIX weenie. And despite the mysterious voices calling out to me from within its briars, I have slim chance of becoming one at this point.

Nevertheless, I do rather enjoy the bent company of those who have mastered it. They're a fun bunch in their propeller-hatted way, though I have increasingly found tedious their disdain for my NeXTophilia. After all, these are mostly guys who think that if God had meant for us to use GUI's, He never would have given us grep, sed, and awk.

But there have been some recurring themes to this criticism that trouble me. Many in the UNIX priesthood seem to believe there is something unpredictably different about the BSD UNIX used by NeXT.

Not long ago, I got tired of arguing with these guys. They are, after all, customers. So I e-mailed Steve Jobs and told him that, in my opinion, these muttering Pharisees were doing great harm to our cause. He suggested I come back with a few details to bolster my case, a request I passed along to my UNIX friends on the Net.

I underestimated the detail with which such an inquiry might be ad-dressed by UNIX zealots. Over the next week or so, I got roughly 20,000 words of response. I didn't know what many of these words meant, nor could I easily decode them based on context. I'd be as qualified to mediate a doctrinal dispute between Jesuits and Discalced Carmelites.

Nevertheless, some complaints were repeated often enough to make one think them legitimate.

1. Everybody agrees on sendmail. NeXT is shipping v. 5.52. The most recent one is 5.65c.

2. UUCP. They're all pretty creaky, of course, but I heard several positive mentions of the HoneyDanBer UUCP.

3. To automount or to autonfsmount. I gather there is some considerable age difference between NeXT's and Sun's. I'm not quite sure what the problem is otherwise, but it kept appearing.

4. More diversity and recent dialer support for tip.

5. NeXT's fingerd and rlogind may have security holes.

6. In order to make the Sun-oids feel more at home, it wouldn't hurt to throw them a few bones like perl, fingerd, and whois, and more GNU-ware, like gcc and gtar.

7. There seems to be a lot of controversy over NetInfo that I found very hard to sort out. This really looks like a cultural problem, but it might smooth a lot of feathers to have something like Sun's flat data files in addition to the files NetInfoManager maintains.

I believe that, with a few such minor adjustments as these, NeXT might even become the preferred platform among UNIX weenies as well as mere mortals.

Lavin: John, the truth is that NeXT's implementation of UNIX is superior to other platforms'. It's just one more example of NeXT planning for the future. For example, NeXT uses Mach as the core of its operating system. Mach is an advanced form of UNIX that was built from the ground up, to facilitate multiprocessing computers. We are just beginning to see the software nightmare as other vendors, such as Sun, try to shoehorn old technology into a multiprocessor environment.

The one area in which companies like Sun and HP do have an advantage is economy of scale. You are completely right that some of the gazillion small programs that make up UNIX are brain-dead on the NeXT. The problem here is the familiar one of NeXT's limited resources. The reality is that the plaintive whining of Joe Geek has little chance to be heard compared with the plaintive whining of Joe Wall Street.

Simon Says
Lavin: I happen to be an amateur expert on the phenomenon of computer-to-human voice communication: I have an '84 Le Baron convertible. It was a luxury auto during that brief shining moment when cars talked to humans. It tells me, due to a faulty sensor, that my washer fluid is low every time I turn on the car. It also tells me, due to a faulty operator, that my seat belt is unfastened and that I have failed to turn my lights off.

On the surface, machines talking to operators have little to do with Simon Says, in which the user talks to the computer. Still, first-time visitors to both my computer and my car are always amused with this spoken communication. A steady diet of either, however, does not even remotely measure up to initial expectations.

Within ten minutes of using Simon Says to voice-control my NeXT, I thought it was the coolest thing ever invented. Within an hour, I had discovered that it was also the first decent macro generator for the platform, and I was even happier. Within three hours, I had an axe through my face from aggravated office neighbors not clued into the joys of loud, mono-tonous voice commands.

But where the talking car concept was totally discredited as a human-interface solution, dying a quick and merciful death, Simon Says is in fact genuinely useful over the long haul.

You start by training Simon with the words you want it to recognize. This involves many repetitions of these words. Since inflection matters in Simon's voice recognition and remembering exactly how you said, "Show Ruler," is next to impossible, you need to develop a consistent style to make the thing work.

Then you can choose which apps you want to have under some voice control. If you are like the rest of us, you'll go hog wild, training everything from PrintManager to BlastApp and all commands from "Insert Soft Hyphen" to "Make Spline."

If you do this, though, you'll quickly find out why applications have menus. A better strategy is to limit the number of trained words to a manageable list covering only key commands in key apps. Once you get the right mix, you'll find that using your voice is truly like having a third hand.

Amazingly enough, Simon also has a powerful macro generator built in. You can construct voice-command macros that will perform keystrokes or primitive mouse events, paste text, run UNIX commands, play a sound, send mail, or do a combination of the above. The resulting macros are so powerful that I wanted them available with keyboard equivalents in addition to voice control.

Like any 1.0 product I like, there are a large number of features I would add or fix. Since voice is a whole new means of interfacing with my computer, deficiencies are magnified. But none of this takes away from the fact that the voice-recognition engine is a thing of marvel. There's no doubt that I'll keep using Simon, though that should come as no surprise. I still drive the Le Baron.

Barlow: Dan, I suppose I shouldn't be surprised that anyone who is "amused" to be prattled at by a 1984 Chrysler Le Baron would also want to spend his days in conversation with a computer. But what the hell, most people talk to their computers all the time anyway and never get the courtesy of a response. Still, the vision of an entire office full of people talking, not to one another, but to inanimate objects, sounds like something out of Brazil to me.

Leaving aside such issues as Modern Alienation, I think Simon Says is pretty cool. It's a lot more successful (if used carefully) than the Voice Navigator on the Mac.

The main solution it provides me is related to the few quibbles I have with both Simon Says and the NeXT. This has to do with mouse religion. I like to lean back out of energized ion range from the monitor and work with the keyboard on my lap. Unfortunately, with a NeXT this means leaning forward a lot to grab for the mouse.

With Simon Says, I can reduce all that stretching a bit. But not much. That's because the macro generator in Simon Says is not very powerful by my standards. Indeed, it works best when keyboard equivalents already exist. Nevertheless, the core feature of the program, its ability to make your computer take orders, just works. I wonder if Simon's developer, Greg Cockroft from Agog, has considered doing a port to 9 year olds . . . .

Weenie wannabe John Perry Barlow and building inspector Dan Lavin mix metaphors with the best of them. Send them e-mail at: and