Thursday, July 2, 2009

Feedback on Economics of LinuxTV

I wanted to take a few minutes and respond to some ideas put forth by Andrej Falout, who was also kind enough to make a donation toward my LinuxTV efforts. Let's take his comments point-by-point.
1) When working on devices where no signal is available, how about using SSH to log in remotely to the box where device is connected and signal such as DVB-T is present?
I actually have done this before (in fact, two weeks ago the Terratec Cinergy T XS USB lock problems were resolved this way thanks to Simon Kenyon providing SSH access to his environment). Sometimes this approach is viable, sometimes it is not. There are significant advantages to having physical access to the hardware, including the ability to use tools like an oscilloscope and logic analyzer. Also, the software environment, antenna configuration, and list of accessible frequencies are known in my environment. To look at another example, I have been working with a different user on the mt352 version of the Terratec Cinergy T XS USB, and I don't know if it doesn't work because of a driver problem of mine, or whether it's user error, poor access to a signal, bad antenna, etc.

For simple things like a new board profile for existing supported chipsets, it can work. For cases where new drivers need to be written or the existing chipset drivers have bugs (which is more often then I like to admit), it is extremely difficult.
2) The key to getting vendors to pay, is to show number of users/devices sold because they work on Linux. Add a message to driver load, such as "Register and rate your device at", as well on the V4L wiki for that device(s). Then when you can show 50 or 100 users registered, contact the vendor(s) and offer to continue supporting device, and maybe even future devices for a reasonable fee...?
There has been considerable discussion of ideas like this between myself and the KernelLabs developers. There are two basic problems: once the driver works, the vast majority of users never even notice (so they have no incentive to seek out the person who did the work), and things like surveys on the v4l wiki only have the opposite effect. They tell the vendor, "there are only ten people who actually care about Linux support for my device, so why should I spent $1000 in consulting for Linux support"
3) may be of interest in determining numbers of devices in use
Again, smolt only hurts the community in this case. Look at those stats. If I were a vendor and I actually took those stats into consideration, they would tell me there might be a few thousand users across the entire industry. I realize that it's a small sample of total users out there, but vendors are not likely to see those finer points.
4) Create few wiki pages on V4L site, for devices you are offering to work on for a fee/donation. Make it clear how much is needed to be contributed for work to start, and provide a easy mechanism for users to donate AND see how many donations are made so far.
Certainly we have been discussing various approaches such as this (surveys, contribution pools, etc). I hope to see some of this come to life in the next few months.
5) Approaching chip makers may be a better choice then talking to the device marketers/assemblers ?
I have worked with a number of chipset vendors, and even getting them to provide datasheets so I can write a Linux driver is usually like pulling teeth (and in reality has only ever happened when I found a product vendor to pressure them to do it). Getting them to go one step further and actually pay for Linux support is even further off.

Some classes of hardware tend to have vendors who are happy to make technical information available (think CPUs). However, the tuner and demodulator vendors are extraordinarily protective when it comes to IP concerns and trade secrets.

Anyway, I continue to look for economic models that are viable. Thanks for helping further the discussion.