Hacker News Reader

Comment by 13throwaway | Hacker News Reader

ELI5 why companies don't open source drivers. The hardware is proprietary so what do they have to lose?

Please don't use things like ELI5, TLDR, and so on, on Hacker News. We strive to write comments in good English.

(Not directed at any comment in particular—there has been a spate of these lately.)

A great deal of performance differentiation in hardware comes from the drivers themselves, not the underlying hardware. This is evidenced in things like differing performance in similar situations using different OSes, such as a game in Windows vs. the same game in Linux (assuming they both us OpenGL).

The techniques that companies use to improve performance on their hardware are almost as complex and important as the hardware itself. This isn't to say that there isn't a benefit in open-sourcing drivers, just that companies are justified in being weary of open-sourcing their drivers.

The differentiation is more in microcode running on the radio chip than in anything that runs on the CPU. The microcode is what chip vendors are really paranoid about; I work on wireless AP firmware, and while we have the source code for all our wireless chips, vendors are very firm about never letting us see the source for microcode. If we need to debug a microcode issue, we file a case with them, and in emergencies they'll send an engineer on-site. We get binary blobs, the driver loads binary blobs, the driver sends config options for the advertised checklist of features to these binary blobs, and that's all we know.

The protectiveness of the driver seems more lawyer-driven. They're willing to hand out the source to pretty much everyone who asks to build a device, just under proprietary licenses.

Even when stuff does get open-sourced, the code takes quite a bit of work to pass muster for e.g. the Linux kernel. It's just so low-quality, regardless of vendor, that a lot of time and effort on the device-maker's end goes into debugging driver issues - work that is of course duplicated in a dozen or more companies. I've never personally looked at Marvell drivers, but OpenWRT's claim that it doesn't meet their standards is completely unsurprising to me.

To continue your point, often companies tier their products by their firmware to save development/manufacturing costs. If anyone can change the firmware, they may be able to make a $99 router perform like a $599 router (if I recall this was one of the big pulls for open-wrt or dd-wrt on the linksys 54g).

This may be true for GPU drivers, but I would argue it is the exception, not the rule.

There is an upper limit of a piece of hardware, but if the software driving it's use is not exposing that or performing actions in a less efficient manner it can certainly have a negative impact.

Belkin effed up their firmware so badly on the F7D4301 that I got a new one for $35.

The original firmware was slow, buggy and required reboots every day. With Tomato it's one of the best deals I ever got :-) - dual band Wifi, Gigabit LAN, support for USB drives and 3G modems, no stupid blinking lights (I love that) - a perfect home/office router.

What tomato distro are you on these days? There seem to be so many I have a hard time figuring out which one looks like it has the longest legs.

I'm using Shibby's RT-N builds. Rock stable and has everything I need, but then again I'm really not picky :-)

I a also guessing that most of the company that sell the hardware do not make it (just repackage it in a different plastic shell or wrap the drivers in an installer with pretty pictures). The hardware making company is likely not that much interested in giving away what it license to ASUS/Apple/Cisco etc....

> is evidenced in things like differing performance in similar situations using different OSes, such as a game in Windows vs. the same game in Linux (assuming they both us OpenGL).

I think better example would be the evolution of performance over different driver versions.

Reply on news.ycombinator.com