Choosing between portability and innovation [LWN.net]

Getting both portability and innovation

Posted Mar 2, 2011 21:59 UTC (Wed) by hamjudo (subscriber, #363) [Link]

I should have written: "While I don't use BSD, I do have many embedded Linux systems that can't run a modern desktop, or in some cases, any desktop. Good monitors cost a lot more than most of those systems anyways, so I sit at the system with the best monitors, and access the development systems using X Windows."

Standards like X11, ssh, NFS, tar, and many more that I don't even realize I'm using, make remote operation easy and painless.

BSD developers who want the latest desktop software can use it, if they want it. It just means that they'll need a copy of Linux running on a real, or virtual host.

Getting both portability and innovation

Posted Mar 2, 2011 22:29 UTC (Wed) by nix (subscriber, #2304) [Link]

You mean the X11 which Wayland is trying to shoot in the head? I'm sure when half your apps are Wayland apps tied to your physical machine that remote work using them will be *such* a lot of fun.

But nobody needs remote access, anyway. That's obsolete, like portability.

Bah.

Getting both portability and innovation

Posted Mar 2, 2011 23:05 UTC (Wed) by einstein (subscriber, #2052) [Link]

> You mean the X11 which Wayland is trying to shoot in the head? I'm sure when half your apps are Wayland apps tied to your physical machine that remote work using them will be *such* a lot of fun.

On the contrary, Wayland will include X11 support - but it won't be loaded by default. That makes sense, since 98% of linux users never use the remote desktop features of X11 - but for those who need it, it will be there. Think of OSX X11 support, done right.

Getting both portability and innovation

Posted Mar 3, 2011 0:06 UTC (Thu) by rgmoore (subscriber, #75) [Link]

The death isn't going to come from lack of X11 support in the system, though. It's going to come when apps are written for Wayland rather than X11. Yes, legacy apps will still work with network transparency, but all the development effort will be done on apps that depend on Wayland-only features. The stuff that works over the network will die a slow death from bitrot.

Getting both portability and innovation

Posted Mar 3, 2011 9:34 UTC (Thu) by roblucid (subscriber, #48964) [Link]

Exactly! And pixel scraping won't work too well, when there's no desktop to scrape pixels off.

Getting both portability and innovation

Posted Mar 3, 2011 0:57 UTC (Thu) by drag (subscriber, #31333) [Link]

Every single day I use remote desktop applications at work. So do most people at my organization.

The clincher is that none of it depends on X... at all.

Getting rid of X as your display server does not eliminate the possibility of using remote applications. Nor does it remove the possibility of using X11 apps either.

It's mostly a red herring when discussing Wayland vs Xfree Server.

Besides all that...

X11 is obsolete, slow, and a poor match with todays technology. It could be something nice, but that would require X12 and if you ever noticed: nobody is working on that.

In fact I think that people are now using more remote applications then they ever did in the past. It's just that relatively few people actually use X11 networking to do it. It's a poor choice for a variety of reasons. I am not describing what I would like it to be.. I just telling it the way it is. The remote access boat has sailed and it's captain is named 'Citrix'.

I would like to change this fact, but X11 networking is not going cut it.

wayland does not preclude application remoting

Posted Mar 3, 2011 1:28 UTC (Thu) by nwnk (subscriber, #52271) [Link]

I grow increasingly tired of this strawman.

No one has defined a remoting protocol specifically for a wayland compositing environment yet. This is good, not bad, because it means we have the opportunity to define a good one. The tight binding between window system and presentation in X11 means that while purely graphical interaction is relatively efficient, any interaction between apps is disastrously bad, because X doesn't define IPC, it defines a generic communication channel in which you can build the IPC you want, which is a cacaphony of round trips.

You want a compositing window system. You probably want your remoting protocol in the compositor. The compositing protocol and the remoting protocol should probably mesh nicely but they're not the same problem and conflating them is a fallacy.

You'll note that wayland also does not define a rendering protocol. In fact it goes out of its way not to. Yet for some reason, wayland is not accused of killing OpenGL, or killing cairo, or killing any other rendering API (besides core X11 I suppose). If anything, that people so tightly mentally bind remoting, rendering, and composition is a tribute to the worse-is-better success of X11.

Unlearn what you have learned.

wayland does not preclude application remoting

Posted Mar 3, 2011 16:38 UTC (Thu) by nix (subscriber, #2304) [Link]

I hadn't thought of putting remoting in the compositor. It seems... odd, but not fundamentally any odder than doing it in the thing which draws the graphics, and I suppose it should work, since the compositor sees everything flowing to and from the user. I suppose you could put all of X11 support in there as well and then not need to support remoting anywhere else. (Which makes me wonder why the X11 compatibility stuff isn't already being done that way... or is it?)