X at OLS [LWN.net]

X at OLS

Posted Jul 23, 2004 3:21 UTC (Fri) by iabervon (subscriber, #722)
Parent article: X at OLS

It's very interesting to see an explanation of how modern windowing
system needs and capabilities fit into the X design. In a world where
people expect to stream full motion video from across the country, it
makes little sense to think you can't use your web browser from across
the room. There are a lot of things that X should do right for the
traditional windowing system devices that it might as well handle for
other devices.

I still feel that it would be beneficial to do an R7 or an X12 which
relegates a lot of old features to separate emulation (e.g., indexed
color, weird XImage formats, etc) and drops them from the new library.

(Log in to post comments)

X at OLS

Posted Jul 23, 2004 11:59 UTC (Fri) by MathFox (subscriber, #6104) [Link]

I'ld suggest to go to an X12 protocol (with only the modern features) and not aim for compatibility with the current X11 implementation. I think that an X11 emulation for X12 servers wouldn't be very difficult; X12 emulation for X11 servers could be, especially if X12 starts using an rgba color model.

X at OLS

Posted Jul 23, 2004 13:25 UTC (Fri) by elanthis (subscriber, #6227) [Link]

Why? After reading this write-up, it should be pretty clear even to someone not familiar with X11 at all that there is *NO* reason to break compatibility. Seriously, what would X12 have different from X11? All you'd do is have a slightly different set of default commands (extensions), and that's absolutely pointless, because you can just do that in X11. The low-level actual protocol of X11 is perfectly fine and doesn't need any changes in order to operate appropriately for modern needs.

X12 (was: X at OLS)

Posted Jul 23, 2004 15:39 UTC (Fri) by MathFox (subscriber, #6104) [Link]

Do we need suport for fixed size bitmapped fonts (only?) 256 colour palettes? The whole set of pixel or/and/not operations?
With a modern graphics card you'ld want antialiassed scalable fonts, an RGBalpha colour model, colour blending, etc. Some of these are in current X11 expensions and should be moved to the X12 core. Some of the current X11 core stuff should be dropped and moved to a compatability library. Keith has some ideas that require deep modifications of the protocol.

X12 (was: X at OLS)

Posted Jul 23, 2004 17:19 UTC (Fri) by bradfitz (subscriber, #4378) [Link]

There's no advantage to having something "in the core". The reason X11 hasn't become X12 is because X11 adapts wonderfully to new extensions as they come and go. You make fun of the bitwise and/or/not operations now, but we'll be making fun of the current state of graphics in another decade.

Why do we want today's cutting edge (tomorrow's crap) "in the core"?

X at OLS

Posted Jul 23, 2004 20:31 UTC (Fri) by iabervon (subscriber, #722) [Link]

I think the main change that should be made is really in Xlib rather than the protocol. A huge portion of the library is stuff that you shouldn't use these days. Another chunk deals with stuff you shouldn't have to worry about. It would greatly simplify X programming if the library insisted on a TrueColor visual and simulated 24bpp on other depths. Of course, cleaning all of this up would require changing header files, so the new version would have to be somewhere other than <X11/...>, suggesting that it would be X12.

It might also be worth making the core X11 stuff an extension for uniformity, and to have a set of standard extensions that you'd be informed of simply by the server telling you what release it is. It might also be necessary to change the lowest-level protocol for some of the virtualization and security stuff.

Please, X11R7, not X12

Posted Aug 5, 2004 16:15 UTC (Thu) by AJWM (subscriber, #15888) [Link]

As somebody who lived through the X10 --> X11 transition, let me voice my support of an R7 rather than an X12. There is still some X10 compatibility code around (see X10.h) and programs that use it -- although many more X10 programs just broke so badly they were completely rewritten. (The original "xtrek" comes to mind as one open example.)

I'm not saying we won't (eventually) need an X12, but that implies major changes to the API. (It could use it -- and I wouldn't mind seeing some high level implementations (C++, etc) that go directly to the protocol rather than through Xlib's semi object-oriented C API.) But there's too much legacy code and legacy hardware out there to completely drop support for it anytime soon. I like that I can use an ancient Sun IPC, an old 486 with a cheapo 2MB SVGA card, or dirt cheap used Pentium laptops as X terminals.

I've got no objection to eye candy per se, I do object to needing a graphic card designed for high-end games just so I can open a few xterm windows and a browser ;-)