:: Re: [Dng] FW: Devuan commitments - …
トップ ページ
このメッセージを削除
このメッセージに返信
著者: Jude Nelson
日付:  
To: T.J. Duchene
CC: dng@lists.dyne.org
題目: Re: [Dng] FW: Devuan commitments - will trade-off be applied?
Hi T.J.,

> *While I am not commenting as to their ultimate aims, which OVERALL are

actually quite admirable and agnostic, it is still less than desirable that
a key technology: KMS IS part of the*
*> initial implementation, which in turn becomes the "reference
implementation" for all practical purposes. KMS as a software is Linux
specific. While there are equivalents out there, they*
*> are certainly less developed. When I criticize Wayland, it is over
concern that Linux-isms have worked their way into the final version of the
stack, making it less portable to systems not*
*> using Linux KMS. I think you can agree that that concern is not
entirely unjustified, given that a number of supposedly platform agnostic
software projects have Linux peculiarities in *
*> their code that make them difficult to port to other UNIXes. *

I don't think I follow. KMS is a generic OS design concept, in the same
way that "memory paging" and "preemptive multi-tasking" are generic OS
design concepts. Each OS that does video mode setting in the kernel has
its own API and implementation for doing so. Maybe one day POSIX will
specify a consistent API that OS developers could implement, but that day
is a long way off.

The BSDs are on-board with not only adding KMS, but also GEM. Both are
much simpler to implement and reason about than trying to get DRM and user
mode-setting to cooperate with X while getting decent performance and good
security. Here's a good perspective from the OpenBSD journal in 2008 on
the subject, for example:
http://undeadly.org/cgi?action=article&sid=20081029164221

-Jude


On Mon, Mar 23, 2015 at 6:36 PM, T.J. Duchene <t.j.duchene@???> wrote:

>
>
>
>
> *From:* Jude Nelson [mailto:judecn@gmail.com]
> *Sent:* Monday, March 23, 2015 3:50 PM
> *To:* T.J. Duchene
> *Cc:* dng@???
> *Subject:* Re: [Dng] FW: Devuan commitments - will trade-off be applied?
>
>
>
> Technically, Wayland is the protocol definition, not the implementation
> (i.e. think X11 vs X.org). Weston is the reference implementation.
>
>
>
> *[T.J. ] Good point. I was using the term Wayland more or less
> generically. Yes, I know, but thank you for the correction. Hopefully,
> it will be less confusing. *
>
>
>
> While it is true that most of the Wayland/Weston developers are also X.org
> developers (and most come from Linux), they are making all the right moves
> to avoid lock-in to Linux or a particular Wayland implementation. For
> example, while there are multiple Wayland implementations (i.e. a Wayland
> window manager or widget toolkit is effectively a Wayland protocol
> implementation), inconsistencies between an implementation and the
> specification are treated as bugs in the implementation. As another
> example, Wayland avoids relying on FreeDesktop technologies like dbus and
> systemd.
>
>
>
> *[T.J. ] Also correct. *
>
> You may be interested to know that the Wayland protocol is not tied to a
> particular rendering technology or paradigm. To use the OSI analogy,
> Wayland lives in the presentation layer--it's concerned with helping
> applications identify and interact with the host's output devices, input
> devices, and pixel memory buffers (on both an individual level and by
> groupings). The protocol itself is not concerned with users, sessions, or
> GPU infrastructure, nor is it concerned with widget sets, windowing,
> decorations, etc.
>
>
> *[T.J. ] Yes, I know.*
>
>
>
>
>
>
> *While I am not commenting as to their ultimate aims, which OVERALL are
> actually quite admirable and agnostic, it is still less than desirable that
> a key technology: KMS IS part of the initial implementation, which in turn
> becomes the "reference implementation" for all practical purposes. KMS as
> a software is Linux specific. While there are equivalents out there, they
> are certainly less developed. When I criticize Wayland, it is over concern
> that Linux-isms have worked their way into the final version of the stack,
> making it less portable to systems not using Linux KMS. I think you can
> agree that that concern is not entirely unjustified, given that a number of
> supposedly platform agnostic software projects have Linux peculiarities in
> their code that make them difficult to port to other UNIXes. I freely
> admit that until we see the final version of Wayland, it is hard to judge
> if it is merely overcaution.*T.J.
>