On Tue, Nov 05, 2013 at 05:56:24PM +0100, Luc Verhaegen wrote: > I personally really dislike this upstream fetish.
It is the only way that has managed to keep things supported long term so far. > The effect is highly detrimental, as it mostly ignores the hard work > needed on getting broken and bad code in difficult places fixed. The > initial and perhaps easy bits make it upstream, and then after a while, > a new platform comes along and interest is lost, and upstream has just > some initial code included, which will only have limited use. At the > same time, the difficult bits will have seen less love as well, and > no-one will have won anything. And then, a few years down the road, > upstream will do what they did with telechips. The first people to loose interest in supporting something is the chip maker in my experience. That is why as both an end user of devices and someone that works with making devices using SoCs, I hate companies that don't aim to mainline things right away. In fact for the next project I am working on we are telling the chip vendors we are talking to that if they aren't working to get things mainlined, then we won't use their chip. Fortunately some of them do seem to already be doing a good job writing clean well documented drivers and working to get the patches accepted upstream. It is too much effort to keep merging patches when trying to move to a newer kernel with new features. A lot of chip makers get stuck with one kernel version, because it does what they think is important, but not what I think is important. If everyone is mainlined, then everyone can actually work together. So it is not a fetish, it is a bloody good idea to mainline. It also means you actually get some comments from really experienced people on your drivers all for free. Why would you not want that? Well unless you think getting the product shipped in 2 weeks from conception is the only thing that matters, never mind how shitty and unmaintainable your code is. -- Len Sorensen -- To UNSUBSCRIBE, email to debian-arm-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20131105171813.gb13...@csclub.uwaterloo.ca
