(formerly)tangentmind.org

So, I was trying to get this SIS 6326(I think that's right...) video card to use Direct Rendering on this old computer I have in my room, in Ubuntu Linux. SIS's video drivers for linux(SIS's video department is now a separate ocmpany called XGI), and subsequently XGI's, really suck.

So one guy, Thomas Winischhofer, out of the goodness of his heart and necessity(he had a laptop with an SIS chipset) wrote a driver for them and has continued to maintain it. While I was looking for some info on trying to get Direct Rendering to work, I happened across something on the forum that he said that I thought was really funny. Someone asked him what someone would have to do to develop a video driver for a video chipset and this was his reply:

"Method A)

1. Find hardware documentation
2. Read hardware documentation
3. Add support for memory size detection, memory bandwidth calculation, mode setting, HWCursor, 2D acceleration, eventually Xv by providing driver primitives that do all this.
4. Spend time with your girlfriend.
5. Be happy.

Method B)

1. Spend days and nights for weeks to figure out how the hardware might work, by writing different values to registers and try to find out what they do.
2. Help your girlfriend to move out of your appartment.
3. Spend further days and nights to figure out how the hardware might work.
4. Eventually implement memory size detection and mode switching to the driver.
5. Be spammed by folks claiming the driver is "dead slow", and demanding futher features.
6: Get stuck with driver development for nearly 5 years.

Method C)

1. Spend weeks with trying to find out how hardware works that you don't even have, by sending out driver binaries to folks and analyzing logs that come back days later.
2.-6. like Method B).

While A) is to be preferred, it often is B) or C) in reality."


It's kind of sad, because it's true, but I think it's funny.