KS2007: Hardware support and the i386/x86

KS2007: Hardware support and the i386/x86_64 merger

Posted Sep 6, 2007 17:50 UTC (Thu) by ckelso (guest, #43128)
In reply to: KS2007: Hardware support and the i386/x86_64 merger by vblum
Parent article: KS2007: Hardware support and the i386/x86_64 merger

Software can be a lot more expensive to change out, for a business, than hardware. This is what gives the developers leverage. A switch to windows (the only reasonable software change you could talk about when drivers are being discussed) is simply not viable for custom business code. If you needed to spend 100k to migrate an application or 4k in new hardware, what would you choose?

At the consumer level, embedded linux is already done in longstanding internal forks against 2.4. They _don't_ have anything to offer the vanilla 2.6 tree. The drivers that they do have are often incomplete or shoddy and aren't available at all to PC users. There is no reason to worry about whether you will chase them off, because they don't contribute even in the binary sense.

(Log in to post comments)

KS2007: Hardware support and the i386/x86_64 merger

Posted Sep 6, 2007 21:00 UTC (Thu) by AJWM (guest, #15888) [Link]

Yep, I've seen that in a couple of cases -- even in cases where it wasn't proven that the hardware/driver was the problem, just suspected. Rip out all the cards and replace them with something we know works (or the 3rd party application vendor supports, whatever).

Replacing a half-dozen fibre cards (HBAs) is cheap compared to what you're paying your SAN vendor, and if he says that card X isn't supported (by him) under the OS upgrade you want to do (or your DB vendor insists on) but card Y is, guess what's going to happen.

(And in this particular case the drivers for both X and Y cards were GPL, so that wasn't even a factor.)

KS2007: Hardware support and the i386/x86_64 merger

Posted Sep 6, 2007 22:06 UTC (Thu) by amikins (guest, #451) [Link]

This doesn't really acknowledge the consumer end, which is where things like binary video drivers generally come into play. Business isn't much of a volatile market for Linux at this time; the major players already know what's offered, and so adoption is pretty much just a matter of vendors getting their sales.

Consumer marketshare is still rather fragile when you look beyond the hobbyists, and consumers tend to develop negative opinions pretty easily. For them, they almost certainly already HAVE another OS which 'just works' with their hardware.. The cost of learning Linux is then added to the cost of new hardware, assuming there is supported hardware that does what they need, and they can easily determine what is supported.

If 'winning' in that space is important, then this is an issue that warrants caution.

KS2007: Hardware support and the i386/x86_64 merger

Posted Sep 6, 2007 23:51 UTC (Thu) by ckelso (guest, #43128) [Link]

There are two markets for consumers as far as linux is concerned. Embedded devices (DVR, routers, network storage, etc...) and low end walmart PCs. Their is simply no demand outside of that. Consumers have _zero_ influence on vendors, sorry. They just don't care what runs on their systems, they only care that what the OS that came pre-installed works. Anyone that does care about running an after market OS is what you call a hobbyist.

Let's look about how hobbyists fair from a kernel perspective. I have never had an issue with a disk controller on desktop box. probably a total of about 10 different chipsets there. I haven't had sound driver issues since ALSA came about. I have had two systems that had XFree86 issues (not kernel per se, but without X I can't say that the kernel works for me either) and which were never solved by binary drivers anyway. I have had two systems that required third party, out-of-tree but still not binary, drivers for network.

Wireless has been and continues to be the biggest issue. The reason it is an issue isn't because of out-of-kernel or binary drivers. The reason it is an issue is because of software radio tuners. Open source software tuners are banned, you just can't make them. Worse yet, the big upgrade chip providers prefer the software tuners because the bulk of the market is a closed driver shop.

All in all though, the binary module issue is a red herring. The drivers that make up the bulk of the out-of-tree drivers available are simply enterprise. This is exactly where the developers have leverage. Why shouldn't they use it?

KS2007: Hardware support and the i386/x86_64 merger

Posted Sep 7, 2007 0:13 UTC (Fri) by amikins (guest, #451) [Link]

>There are two markets for consumers as far as linux is concerned. Embedded devices (DVR, routers, network storage, etc...) and low end walmart PCs. Their is simply no demand outside of that. Consumers have _zero_ influence on vendors, sorry. They just don't care what runs on their systems, they only care that what the OS that came pre-installed works. Anyone that does care about running an after market OS is what you call a hobbyist.

I suspect there's little I can do to sway your opinion based on these statements, but I think it's worth noting that there's an increasing tendency for people to notice the sorts of things they're getting put through on their attempt to keep 'current' on their technology. People who are most definitely consumers -- not hobbyists by any stretch of the imagination -- are noticing that XP isn't coming on new computers anymore; Vista has started being noticed, and some people aren't liking it. This is causing regular people to start looking more and more at other options.

I've had several individuals who historically would have been the 'well, it works, so whatever' type ask me if I know anything about this 'Linux' thing, because they don't like Vista and want to know if they can get their work done on something aside from what Microsoft is pushing. There isn't significant marketshare CURRENTLY occupied by Linux for consumers.. But there's potential of it. The end-user class of software is pretty close to ready.

But there's a risk of a self-fulfilling prophecy. If something gets written off before it has a chance to happen, then that only ensures that it won't happen.