|
1000th reply
|
#
?
Jun 14, 2014 00:29
|
|
|
|
| # ? Jun 8, 2023 19:16 |
|
|
Notorious b.s.d. posted:are there any other (current) consumers of your xdg-shell work? EFL and Qt are ported to it. SDL2 maintainers are porting to it right now. Weston and mutter both support it natively, while Orbital and the Qt compositor have branches.
|
|
#
?
Jun 14, 2014 00:50
|
|
|
hmm, i was rather hoping for more meat on xdg-shell vs wl_shell and the nature of its associated deficiencies. is there a ML thread somewhere at least where this is discussed?
|
|
#
?
Jun 14, 2014 01:18
|
|
|
Nomenclature: Wayland is a message passing protocol. The messages from a client to a compositor are called "requests", and the messages from a compositor to a client are called "events". So, wl_shell_surface is architected with the client in control of a lot of things. The client controls when the window is maximized, and that means that there's no way for the compositor to have a keybinding or gesture (drag to top of the screen) to maximize a window at all. We could add additional events to that to say "hey, client, please maximize this surface", but that comes with its own problems. For a tiling WM, you really always want the "maximized" state, too, so the concept of the client controlling everything isn't great. So, really, what you want is a "hey, client, you're maximized now" event. So, there's a wl_surface.attach "request" to attach a new buffer to the surface. If an app is redrawing at 60fps, then you'll call wl_surface.attach 60 times a second. The issue with the "hey client, you're maximized now" event is that there's a race condition when transitioning states. If we do the maximize binding and the client has an attach in-flight, we aren't sure whether the client has noticed the "you're maximized now" event, and so we have no idea if the attached buffer is for a maximized client or not. So, the client is rendering at 60fps. It makes an "attach", and it's queued up in the compositor's socket buffer. The compositor sees a keybinding, it sends "hey, you need to be maximized", and then notices the attach in its buffer. It is going to move the surface to the top-left, since it's expecting the maximized buffer to be full-sized, and what you end up seeing is the surface be in the top left corner but not have the full size of the monitor. It looks awful. xdg-shell solves this by having one event that tells the client the size that the compositor requests it should be, along with a bitfield of states that the client is in (maximized, fullscreen, among others). This is known as the "configure" event. It also makes the client send a sentinel request, "ack_configure" immediately before it attaches its new buffer so the compositor knows where it in the stream. When the compositor sees the "ack_configure" event, it knows that the client has seen the event, and so it goes ahead and does the full state transition with the new buffer. So everything is atomic and fluid: you hit the keybinding, and it maximizes without any intermediate state. It's all the small touches like these that complicate the protocol, but make the experience more fluid and modern.
|
|
#
?
Jun 14, 2014 01:37
|
|
|
ah neat, thanks for the explanation. i look forward to trying out this smooth glitch-free wayland stack sometime soon
|
|
#
?
Jun 14, 2014 01:40
|
|
|
Suspicious Dish posted:Nomenclature: Wayland is a message passing protocol. The messages from a client to a compositor are called "requests", and the messages from a compositor to a client are called "events". you should add this to the article the concrete example really illuminates the value
|
|
#
?
Jun 14, 2014 04:35
|
|
|
I was going to add something like that, but figured it was already getting a bit too long, so I just cut it there. It's already out to the blog aggregators. I'll explain a bit more in a followup blog post next week I guess.
|
|
#
?
Jun 14, 2014 04:42
|
|
|
im ga.y
|
|
#
?
Jun 14, 2014 04:42
|
|
|
Bloody posted:im ga.y funny.computer!
|
|
#
?
Jun 14, 2014 04:45
|
|
|
Suspicious Dish posted:funny.computer! dont post my email provider
|
|
#
?
Jun 14, 2014 04:46
|
|
|
Here you go Mr. Dog: http://afaikblog.wordpress.com/2014/06/18/a-notifications-update/
|
|
#
?
Jun 18, 2014 18:32
|
|
|
This thread is real cool. I do a little bit of halfassed Linux sysadmin at work and run a lubuntu laptop but I like knowing what's behind the scenes if you will
|
|
#
?
Jun 18, 2014 23:41
|
|
|
Captain Foo posted:This thread is real cool. I do a little bit of halfassed Linux sysadmin at work and run a lubuntu laptop but I like knowing what's behind the scenes if you will more like whats behind the curtains
|
|
#
?
Jun 18, 2014 23:49
|
|
|
E! True Linux Stories
|
|
#
?
Jun 19, 2014 00:10
|
|
|
api call girl posted:more like whats behind the curtains more like if the curtains match the drapes
|
|
#
?
Jun 19, 2014 02:30
|
|
|
The Management posted:more like if the curtains match the drapes they're both made out of beard hair, so, yes.
|
|
#
?
Jun 19, 2014 02:39
|
|
|
i had an iptables question while the forums were down but rose to the occasion and figured it out myself. radium moves in mysterious ways.
|
|
#
?
Jun 26, 2014 04:39
|
|
|
hey suspicious dish i thought of a use case of mine that could actually be filled by wayland, if people are thinking about it sooner rather than later: multiple DPIs, different DPI per monitor so loving sick of plugging in an external monitor with half the dpi of my desktop and having X11 render everything exactly the same on both displays even just scaling the painted windows, leaving the clients unaware, would be better than the current state
|
|
#
?
Jul 16, 2014 13:20
|
|
|
That's a use case that Wayland can support, yep, and Weston can support it already, I believe.
|
|
#
?
Jul 16, 2014 15:38
|
|
|
for once i do not dread the grim future of the linux desktop
|
|
#
?
Jul 16, 2014 21:33
|
|
|
is lunix on the desktop arrived yet
|
|
#
?
Jul 16, 2014 22:37
|
|
|
which loonix do i put on my laptop
|
|
#
?
Jul 16, 2014 22:37
|
|
|
Bloody posted:which loonix do i put on my laptop fedora
|
|
#
?
Jul 16, 2014 22:39
|
|
|
the entire linux movement have always been incredibly stupid randomly replacing things that more or less works for purposes of ~purity~, ~~vision~~, and, god loving forbid, ~~~good engineering practices~~~ (see also: linux audio not working at any point in the past 20 years). but wayland is soooo fuuuuuuucking loooooooooooooong overdue x11 was poo poo the day it was designed, wayland appears incredibly sane in comparison, and in fact a lot of the toolkit and similar work is pretty drat decent but hampered by a hilariously bad underlying infrastructure. trust loving nerds to fetishize the pointless network transparency as a too desirable feature to give up, at a point where every other os had far superior network transparency by copying dumb byte buffers around
|
|
#
?
Jul 16, 2014 22:40
|
|
|
also, linux on the desktop is never happening but at least linux will perhaps stop being an incredible nuisance in the niches where it works if the graphics stack situation (well, and audio and such) can get cleaned up. if you are a 90%+ market share montrosity like windows you can get away with some cruft and annoyances, but linux having a far worse graphics stack was always a huge sticking point even for the niches where you can pick whatever
|
|
#
?
Jul 16, 2014 22:43
|
|
|
emptyquote if ur still not using lunix on the desktop
|
|
#
?
Jul 16, 2014 22:44
|
|
|
theadder posted:emptyquote if ur still not using lunix on the desktop
|
|
#
?
Jul 16, 2014 22:45
|
|
|
wait does a development vm that i use once a month count
|
|
#
?
Jul 16, 2014 22:45
|
|
|
nope
|
|
#
?
Jul 16, 2014 22:45
|
|
|
theadder posted:emptyquote if ur still not using lunix on the desktop
|
|
#
?
Jul 16, 2014 22:45
|
|
|
wait, anyone runs linux on the metal in any other role than as a virtual machine host in 2014? x11 sucks in vm setups too~~
|
|
#
?
Jul 16, 2014 22:46
|
|
|
Cybernetic Vermin posted:x11 was poo poo the day it was designed, wayland appears incredibly sane in comparison Wayland could not have been done 30 years ago when X11 was invented. You didn't have enough memory for buffer contents for every window. Not to mention that every app rendering into buffers independently is really dumb when you have limited memory. The design of X11 of "a server that hands out pieces of the front buffer" was a smart design for the goals that the authors had in the day: a low-memory, fast graphics system that could work across multiple terminals so that expensive apps could be used in multiple places. The rapid increase in available memory, increase CPU speed, and development of a new pixel processing architecture simply means that the goals of X11 became obsolete. If I were to design a system that tried to meet the same goals, I'd end up very close to X11. I don't hate X11 for trying to meet its goals, which were very good goals back in the day. I hate how we've piled poo poo onto it over the years, flailing to adapt the architecture to meet new goals, without realizing when these goals conflict. Cybernetic Vermin posted:and in fact a lot of the toolkit and similar work is pretty drat decent but hampered by a hilariously bad underlying infrastructure. trust loving nerds to fetishize the pointless network transparency as a too desirable feature to give up, at a point where every other os had far superior network transparency by copying dumb byte buffers around I have no idea what "network transparency" means. I couldn't add it to Wayland even if I wanted to, because I have no idea what the goal even is. Usually it's used for a replacement for "remote display" like with ssh -X, but that's better done by a proper remote display protocol like RDP. Hell, people talk about NoMachine's NX as being a perfect example of X11's "network transparency", but that protocol isn't X11, it's actually a lot more like RDP than "remote rendering". NX's software actually often client-side rendering and ships bytes across the wire rather as video-codec-like compression rather than pushing rendering commands to the remote server.
|
|
#
?
Jul 16, 2014 23:04
|
|
|
better things than x11 were designed before x11, wayland is a *lot* better, but x11 was total poo poo the day it was invented. it is sort of a classic idiot nerd thing to hail old disfunct poo poo as good, but NeWS predates x11, and *really* *was* *better* in *every* *way* network transparency is a standard term in reference to the x11 protocol and i am rather perplexed how i managed to introduce it to a redhat developer. you more or less make the intended point; relatively dumb but lower-level byte transport works a lot better than what x11 offers natively
|
|
#
?
Jul 16, 2014 23:19
|
|
|
No, I've certainly heard the term "network transparency" before, and sometimes by coworkers. It still doesn't describe any set of achievable goals, nor why traditional remote desktop solutions like RDP and VNC don't count, but extremely similar protocols like NX do.
|
|
#
?
Jul 16, 2014 23:46
|
|
|
Suspicious Dish posted:No, I've certainly heard the term "network transparency" before, and sometimes by coworkers. It still doesn't describe any set of achievable goals, nor why traditional remote desktop solutions like RDP and VNC don't count, but extremely similar protocols like NX do. theres a video by a the dudes behind wayland where he basically explains that x11 is awful at being network transparent also lol@people complaining about plan9 being naive about network transparency and then clinging to x11 lovely network transparency
|
|
#
?
Jul 16, 2014 23:52
|
|
|
Suspicious Dish posted:No, I've certainly heard the term "network transparency" before, and sometimes by coworkers. It still doesn't describe any set of achievable goals, nor why traditional remote desktop solutions like RDP and VNC don't count, but extremely similar protocols like NX do. b/c vnc is awful, rdp is a lot better but still sucks, and sperglords still think that X11 is fast because they used it over lan links and assume its fast over wan links there isn't a good solution out there.
|
|
#
?
Jul 16, 2014 23:58
|
|
|
Malcolm XML posted:b/c vnc is awful, rdp is a lot better but still sucks, and sperglords still think that X11 is fast because they used it over lan links and assume its fast over wan links the problem with x11 is not that x11 is terrible, it's that client-side fonts are a disaster. if you build rxvt or emacs without client side fonts, they're great across wan links. client-side fonts basically turn x11 into a really bad, confusing version of vnc
|
|
#
?
Jul 17, 2014 00:12
|
|
|
theadder posted:emptyquote if ur still not using lunix on the desktop begging for emptyquotes is just sad it's like asking for a buyback
|
|
#
?
Jul 17, 2014 00:13
|
|
|
Cybernetic Vermin posted:better things than x11 were designed before x11, wayland is a *lot* better, but x11 was total poo poo the day it was invented. it is sort of a classic idiot nerd thing to hail old disfunct poo poo as good, but NeWS predates x11, and *really* *was* *better* in *every* *way* X predated NeWS, and NeWS was proprietary. there are a thousand forgotten proprietary window systems. i'm not sure why people eulogize NeWS and forget about att 3b2 or sunview and all the rest
|
|
#
?
Jul 17, 2014 00:14
|
|
|
|
| # ? Jun 8, 2023 19:16 |
|
|
I, too, as a software engineer, have opinions on NeWS. They do not hold NeWS very highly regarded. It's a similar system to Plan 9's rio, and I don't like either architecture. I won't go into detail.
|
|
#
?
Jul 17, 2014 00:17
|
|