Why commercial Linux software ventures fail

– by Robin “Roblimo” Miller –
I was so excited when my review copy of Corel’s (then) new WordPerfect Office for Linux came that I almost ripped the box apart in my eagerness to stick the CDs in the drive and start using this well-regarded office suite. The only problem was that it wouldn’t install without so much fiddling and patching that I screamed many obscenities in the general direction of Corel’s corporate office. And Corel is not the only company that has tried to sell commercial Linux software so hard to install and use that no sane person would want to spend money for it.My personal distribution preference is Mandrake. This doesn’t mean it’s better than all the others, just that I am used to its quirks. I also prefer a KDE desktop. Again, this is a matter of personal taste, not an indication of technical superiority. But Mandrake and KDE are popular choices. If you are selling a user-level Linux application, you need to make your product work with Red Hat, SuSE, Red Hat, and Caldera, because these are the four distributions the vast majority of non-hacker Linux users are likely to have installed on their desktops or laptops, and among the Linux-for-office-tasks crowd, KDE is far more popular than Gnome, although Gnome is popular enough that compatibility with it is necessary for wide acceptance.

Back to the WordPerfect experience: The installer hung on a “font manager” package. Okay, maybe it was a Mandrake problem. I had another desktop around with Red Hat on it. Same problem. I emailed Corel, and got an email reply telling me what I had to do, namely add a bunch of code to various files and delete some from others. There were no clear instructions on how to find those files or where the additional code should be added into the hundreds of lines they contained. A friend of mine who is an experienced Linux sysadmin tried to help me, and he found this series of script modifications tedious, too. In the end we got WordPerfect installed and working with Red Hat. And then it crashed. Repeatedly. More email to Corel. More lines we could try modifying.

Finally we tried Corel’s own Debian-based Linux distribution, which had a nice and smooth install on the simple desktop we put it on, and got WordPerfect to install without a lot of dancing. That gave us a desktop with exactly one working user-level application: WordPerfect Office. All the others we needed would have had to be downloaded and installed separately, and even though apt-get is a wonderful tool, this is still time-consuming. And then there would have been all the work of archiving work files and reinstalling them in the new system. I was looking at a minimum of two days solid work to completely move from Mandrake to Corel’s version of Debian just to use WordPerfect Office, and it still crashed too frequently to be practical for everyday use when running on its own distribution. End result: no kind words from me about WordPerfect Office.

Then there was ViaVoice. As much as I write, nothing would make me happier than being able to talk into my computer and have text appear on the screen, even if I had to edit that text like mad before it was usable. Speech-to-text is my personal killer app, the one I would love to have above all others. So I bought the commercial version of ViaVoice directly from IBM and tried to install it. Naturally, the Java Runtime Environment I had was the wrong one, so I had to install the one ViaVoice needed. Then I found that the headset supplied in the box with the software didn’t fit the connectors on the laptop I was using for my daily computing needs, so I went and bought a better headset.

At this point I had ViaVoice actually installed, and was trying to perform the “registration” procedure, essentially training the software to recognize my voice. But I kept getting error messages, or nothing at all, no matter how many times I read the IBM-supplied “test” paragraphs into the microphone. I joined an email discussion list IBM had set up to help users solve ViaVoice problems. This was how I learned that ViaVoice probably wouldn’t work on a laptop at all, and that it probably wouldn’t work with stereo sound cards. The recommendation was USB sound, which comes into the computer as digital signals instead of as interference-prone analog. Okay. I bought a USB sound adaptor. Fiddled around to get that to work. Tried again with ViaVoice. Still no-go. Helpful IBM people kept putting out email with more tricks to try, tweaking this and changing that, but by this time I was out at least 20 hours and $300, so I gave up and accepted the fact that ViaVoice worked only for IBM employees at trade shows, not for ordinary people.

To be fair, I must point out that almost all voice-to-text applications are hard to set up and get going, and that most people I know who have bought them, no matter what operating system they use, end up disappointed. And Corel and IBM are far from the only offenders in the Poor Linux Software Value category. I have had similarly poor experiences with “pay for” Linux programs from at least a dozen other vendors.

Last week I wrote about how commercial apps for Linux need to be enough better than their free competition to justify their cost. The two examples above simply weren’t worth paying for. If they had been given away free, clearly labeled as hobby items, no one would have been upset. But they weren’t. They were sold as finished applications, with sales copy that led user-level customers to believe they were buying something that would install and work, right out of the box.

No one says it’s easy to write “click and go” install scripts that will work with all popular Linux distributions and, at least, Gnome and KDE desktops. But any commercial Linux application that is going to generate enough sales to be worth writing must have this capability, and it must be tested thoroughly before the first “look at our new product” press release is sent out.

And, of course, once the product is installed, it must work as advertised. This is professionalism. And the lack of it in so many commercial Linux applications is a big reason why Microsoft employees and apologists get nods from mainstream trade show and conference audiences when they say Linux is still a “hacker toy” with a high TCO (Total Cost of Ownership) that is not ready for widespread adoption.