08-23-2012, 06:35 PM
If that's the case, then I really do wonder what your forum posts are.
Originally Posted by BO$$
08-23-2012, 07:14 PM
I think this thread has had enough ranting already. Your point of view as an X developer would be a welcomed addition.
Originally Posted by daniels
08-24-2012, 04:42 AM
Ha. The tag line's actually a little inaccurate, since I've been working completely on Wayland all year, and haven't really done much for X since (except revive and merge some old patchsets).
I think Wayland's really good, and has a lot of promise. It's a fundamentally different approach to X in a number of ways: the server/compositor split, having the interfaces be descriptive rather than prescriptive (e.g. 'this window is a popup', rather than 'place this window exactly here and give me all input until I tell you otherwise' - which is the reason you still can't use media keys or have your screensaver trigger with a popup open in X in 2012), the concept of a frame being first class, etc. A lot of Wayland was strongly inspired by a lot of extensions to X (DRI2, Xi2, et al), but the reality is that there are some things in core X that we just can't fix with extensions, ever.
X isn't a bad system, not at all. But X was designed for 1985, not 2012, and the problem now is that there's just a huge mismatch between what modern toolkits and apps try to do, and what X lets you do, that everything we've been doing for the last few years has been trying to get around X's core design principles. Some of those were quite successful, some of those less so, and some of those work fine but at a huge cost. If you look at how X is used these days, mostly it's as a very expensive IPC mechanism between clients and the window manager/compositor, trying to very painfully synchronise every step, and also trying to work around things like X trying to automatically draw for you. Very, very little of X is actually used.
Of course, X isn't going to die: not this year, not next year, not in 2015. It's always going to be there, and that's one of the reasons why XWayland is being pushed as a first-class integrated solution. I'm sure it's got some life in it, but we've now got to the stage where it takes a gargantuan amount of effort within X to make very slight improvements, and personally I feel that I'm better off putting that effort into Wayland, where we can see pretty huge improvements for much less effort. Working in the mobile space as well, a lot of our customers have UI requirements which we simply can't ever fulfill with X - at least not without #ifdef'ing chunks out of the server and making it technically non-compliant. That, and I've been working on X for ten years, and they say change is as good as a holiday.
That's my opinion, anyway. I don't buy into all the polarised tribal bullshit, nor the notion that it's a zero-sum game and one must improve to the detriment to the other. Wayland exists precisely because of what we've all been doing within X.Org for the past few years (a lot of it with the express goal of making it easier to develop other window systems, on the grounds that the total lack of competition was deeply unhealthy for the whole system), and as many people have noted before, a lot of Wayland contributors (including Kristian himself) have long backgrounds as X developers. We all still go to the same conferences, work together on the same lists and the same code, and share the same goal of pushing open source window systems further.
As for AMD and NVIDIA, well, they write the blob for people who buy enterprise/industrial workstations by the tens of thousands of GPUs, not for the Phoronix forums. So until those customers start demonstrating interest in Wayland, then I wouldn't expect to see anything in the blobs.
08-24-2012, 04:57 AM
Well, there's a nice informative post, thank you very much.
If only it could be made into a sticky. Or maybe Michael can post a link to this in his every article about Wayland