Ubuntu Allegedly To Have Its Own X, Wayland Alternative

[phoronix]

Phoronix: Ubuntu Allegedly To Have Its Own X, Wayland Alternative

There's been talk already this morning in the forums, Twitter, and via email to Phoronix that Canonical is allegedly developing its own display server rather than using X.Org/X11 or Wayland...

http://www.phoronix.com/vr.php?view=MTI5MjI

[silix]

<...> So just "taking out the old stuff" does not necessarily work. Certain parts of the stack WOULD have to be redesigned and rewritten to make sure the code is clean at the end.

exactly

it wouldnt work nor be enough, when looking at the at the "X12" proposal (or the "Why X is not our ideal window system" paper), you realize that certain aspects of X are "less than ideal" (when not straight out broken) at the core protocol level (like the 16 bit unsigned dimension / 16 bit signed coordinate model allowing for 32K x 32K pixmaps with 3/4 of their surface unaddressable - or the drag and drop and clipboard model..)
and that if you want a sane and modern infrastructure (especially, one accomodating GPU's as first class citizens) what remains after taking parts away also need to undergo significant rethinking ( at which point it becomes something which is not X any more...)

[Ericg]

Then so be it.
Just strip out everything that can be stripped out, and if all GTK and Qt applications still work, and almost nothing of X is left, then that is just good, right?

Then the codebase will be smaller, it will be easier to maintain, quicker to compile, easier to fix bugs, and everything will be great.

There's one big impediment to X11-Light that does shout back to my issue of design it out, spec it out, write it out. And Daniel hit on it during the talk, X has 3 API's for input...and they are all inter-dependent on eachother. It frightens me to think about other spots in the Xorg stack where we have a similar issue where API's are inter-dependent on eachother AND redundant of eachother. So just "taking out the old stuff" does not necessarily work. Certain parts of the stack WOULD have to be redesigned and rewritten to make sure the code is clean at the end.

[Ericg]

The worst was whatever Citrix ICA used. I think they rolled their own scheme. That thing was unusable.

Never used it, probably for the better lol

[johnc]

Compared to what? VNC? No, I'll take RDP over VNC any day.

The worst was whatever Citrix ICA used. I think they rolled their own scheme. That thing was unusable.

[Ericg]

That bad, eh? Shit.

Compared to what? VNC? No, I'll take RDP over VNC any day.

[KellyClowers]

We may finally get Windows RDP quality

That bad, eh? Shit.

[JS987]

Yes, and those applications should still work too.
But there are lots of other cruft that we don't need to support, such as glyphs and drawing operations.

Text rendering shouldn't be removed as it is accelerated by graphics card.
rxvt-unicode won't work without XDraw* functions.

[Ericg]

As I understand it Xorg had when it was as biggest almost a million lines code, if I remember phoronix article right wayland has around ten thousands line ( in the middle of 2012).
Xorg light need to be very light if it should get that small. How big is the refactoring work that is needed to do something like that?



By the way, its look like Arch linux has wayland in their regular package repositories now.

There about, Daniel's video on Xorg -> Wayland quoted 900k though he said at one time he could've sworn that he saw 1.2Million. A lot of that was because in order to be platform-agnostic X had to do a lot of things that are now handled by other programs. For example, once upon a time Xorg had an entire printer server & stack built in because at the time none of the unix's had a standarized way to do printing, now we have CUPS and Samba but at the time it was needed. (Yes that got removed awhile back, Daniel said it was the largest, single, code purge in X's history lol)

[Ericg]

I think the biggest weakness regarding Linux display services is the fact that proprietary video drivers are almost an afterthought. It seems that Wayland is going to continue the tradition, though I guess most of it is license related so there's not much that can be done anyway.

But Android pulled it off pretty well.

And maybe this is what Canonical is focusing on with their own implementation (should it come about), because they know they'll finally have to get real with video drivers if they want to enter the phone market.

Yes and no. DRI3 is working on allowing multiple OpenGL implementations to be installed at once so you could have both Radeon and Catalyst installed at once and switch between them with a reboot. There's nothing in Wayland that says "No closed source drivers!" All wayland does is stake pointers and buffers and displays their contents. Anything BEFORE the buffer, such as how that buffer got filled or who wrote to it, is a client problem with the one exception that Wayland will not allow programs to write to (or read?) eachother's buffers. If you need to share buffers you need to do shared memory or ideally, DMA-BUF