The Gimp seems like a perfect project to found a Cygnus-type company around (or perhaps Apache with IBM is the better analogy). Rather than having companies such as those mentioned sponsoring individual developers, why not sell support/consulting/development services to those companies? One could also sell branded, tested, supported releases a la Red Hat.
If the software development world really is going to change radically towards open source development, these sorts of businesses have to start popping up.
I've used laptops as my primary computing platform for most of my career, and I've run into the same sorts of problems mentioned in the article running anything that wasn't Windows 3.1 (in the early days), or Window 95. Even as of a couple years ago, one had a limited set of options if one wanted to purchase a laptop on which NT was officially supported. (It is probably still the same).
The problem is not only the WinModem evil: Laptop hardware seems to be such a custom beast, that sound and video drivers need to be custom written. I remember trying to get NT 3.5 and 3.51 to run on one of the first Pentium laptops: that was a worse configuration effort than anything I have seen with Linux. Beta this, alpha that, to get any sound, and to get anything other than the lowest resolution.
My primary computing platform is an older Tecra running RH 6.0. I had to tweak some XF86Config options to get proper resolution, and compile a kernel to use the proper sound card settings, but it was all standard Linux. By contrast, to get NT working, I would have to download a set of custom video and sound drivers from Toshiba. So Linux is actually in a better situation than NT here.
Do you make your money off of hardware, support, consulting, and add-on software, or on your proprietary OS?
Are there unique strengths your company has? (IBM=global enterprise services, SGI=visualization, Sun=network is the computer)
In the short or medium term, are you shifting to an Intel hardware strategy (ia32, ia64)? (SGI, HP, IBM?)
In the short term, can you live with less than enterprise level? (SGI)
Why spend oodles of money developing the OS whose only function is to create a market for your hardware, services, unique value, if you can spend less money developing those features into Linux? In the final analysis, your competitive strength is not your OS (especially SGI, IBM, HP; Sun seems to be more tied to Solaris), and it is defocusing to pretend that the OS does anything more than help you get into the market where you strengths show.
The problem for the moment for IBM, Sun, HP, is that Linux is seen mostly as an Intel solution, and that it doesn't scale up to enterprise levels. But it makes perfect sense for SGI to phase out Irix in favor of a Linux with all the support for visualization that plays to SGI's strengths. And it makes sense for the rest to shift R&D from proprietary OSs, into Linux, to develop the enterprise level features and strength on non-Intel platforms to allow them to phase out AIX, HP-UX, Solaris, etc.
This is a clear win for Linux (witness XFS, etc.), and we ought to encourage this as much as possible.
Having been involved in commercial software development, I think I can explain the phenomena. Many people don't appreciate that quality assurance (QA) and maintenance are at least as costly as actual development, if not more so. (This is one of the areas where bazaar-style development helps quite a bit: it is much more efficient to have hundreds of users hitting a software package with something approximating actual usage, than to have tens of QA engineers hitting the software with test cases).
One of the services ISVs such as Red Hat provide is quality assurance. When you purchase a production distribution, you expect to have some assurance that all of the packages provided at least play well together. As anyone who has tried to upgrade software packages can well understand, getting all the dependencies to work well can be non-trivial. Since no two distributions have the same versions of all the packages, and as most software, commercial or non, relies on multiple packages, one can well imagine the benefits of targeting only one distribution, if you are involved in non-bazaar-style development, and especially if you are are developing closed, binary only software. Targetting multiple distributions causes at least a linear increase in QA and support effort, and the increase can be exponential (imagine 2 or 3 tier client/server products, where client 1 + server 1 may test differently from client 1 + server 2, etc.)
If you have to choose one distribution, you choose the most popular.... The "supports Red Hat" line is generally an indication that it was only tested on Red Hat, and that the support and maintenance staff only have Red Hat installed to reproduce reported bugs, and the like.
Slashdot Mirror
If the software development world really is going to change radically towards open source development, these sorts of businesses have to start popping up.
The problem is not only the WinModem evil: Laptop hardware seems to be such a custom beast, that sound and video drivers need to be custom written. I remember trying to get NT 3.5 and 3.51 to run on one of the first Pentium laptops: that was a worse configuration effort than anything I have seen with Linux. Beta this, alpha that, to get any sound, and to get anything other than the lowest resolution.
My primary computing platform is an older Tecra running RH 6.0. I had to tweak some XF86Config options to get proper resolution, and compile a kernel to use the proper sound card settings, but it was all standard Linux. By contrast, to get NT working, I would have to download a set of custom video and sound drivers from Toshiba. So Linux is actually in a better situation than NT here.
- Do you make your money off of hardware, support, consulting, and add-on software, or on your proprietary OS?
- Are there unique strengths your company has? (IBM=global enterprise services, SGI=visualization, Sun=network is the computer)
- In the short or medium term, are you shifting to an Intel hardware strategy (ia32, ia64)? (SGI, HP, IBM?)
- In the short term, can you live with less than enterprise level? (SGI)
Why spend oodles of money developing the OS whose only function is to create a market for your hardware, services, unique value, if you can spend less money developing those features into Linux? In the final analysis, your competitive strength is not your OS (especially SGI, IBM, HP; Sun seems to be more tied to Solaris), and it is defocusing to pretend that the OS does anything more than help you get into the market where you strengths show.The problem for the moment for IBM, Sun, HP, is that Linux is seen mostly as an Intel solution, and that it doesn't scale up to enterprise levels. But it makes perfect sense for SGI to phase out Irix in favor of a Linux with all the support for visualization that plays to SGI's strengths. And it makes sense for the rest to shift R&D from proprietary OSs, into Linux, to develop the enterprise level features and strength on non-Intel platforms to allow them to phase out AIX, HP-UX, Solaris, etc.
This is a clear win for Linux (witness XFS, etc.), and we ought to encourage this as much as possible.
One of the services ISVs such as Red Hat provide is quality assurance. When you purchase a production distribution, you expect to have some assurance that all of the packages provided at least play well together. As anyone who has tried to upgrade software packages can well understand, getting all the dependencies to work well can be non-trivial. Since no two distributions have the same versions of all the packages, and as most software, commercial or non, relies on multiple packages, one can well imagine the benefits of targeting only one distribution, if you are involved in non-bazaar-style development, and especially if you are are developing closed, binary only software. Targetting multiple distributions causes at least a linear increase in QA and support effort, and the increase can be exponential (imagine 2 or 3 tier client/server products, where client 1 + server 1 may test differently from client 1 + server 2, etc.)
If you have to choose one distribution, you choose the most popular.... The "supports Red Hat" line is generally an indication that it was only tested on Red Hat, and that the support and maintenance staff only have Red Hat installed to reproduce reported bugs, and the like.
Yes, but did you notice the newspaper headline on the eve of 1999?
"Year 2000 here, doomsayers upbeat"