wayland-protocols scope and governance

Thu Feb 21 17:11:46 UTC 2019

On Tue, Feb 19, 2019 at 04:50:27PM +0000, Daniel Stone wrote:
> Hi all,
> I'd like to open up a discussion on enlarging wayland-protocols to a
> wider audience, with a better definition of what it contains.
> 
> Currently, wayland-protocols is a relatively small set of protocols
> which were either grandfathered in from Weston, or a semi-opinionated
> set of protocols that someone thinks is 'good'.
> 
> The original intent was to provide a set of 'blessed' desktop
> protocols which 'everyone' would probably implement in order to
> provide a coherent Wayland environment. To some extent - xdg-shell,
> dmabuf, xdg-output, viewporter - this succeeded. For some others, it
> failed badly - the input protocols no-one likes or has implemented,
> necessitating Dorota's rewrite.
> 
> The elephant in the room is extensions like layer-shell and some of
> the related extensions to build a desktop environment from a set of
> disparate clients using generic APIs. Personally I think the
> experience of X11 shows it's only designing for pain, and this is the
> general position of wayland-protocols at the moment. But on the other
> hand, those protocols aren't going away, they are in use, and having
> them developed in a separate siloed community is doing us all a
> disservice, since neither hand fully knows what the other is doing.
> Even if we don't agree on the fundamentals of the protocol, we could
> at least discuss it and try to point out some pitfalls and make some
> suggestions for improvement.
> 
> A related issue is that it's hard for both application and compositor
> authors to figure out what to do. There is no good 'big picture' on
> how these protocols fit together, nor can people figure out which of
> the competing proposals they should be using if they want to write an
> application running on a given compositor, nor can compositor authors
> figure out what apps want them to support. Depending on who happens to
> be paying attention to the particular forum the question is asked,
> they might get very different answers, depending on the point of view
> of who answers.
> 
> My first, hopefully uncontroversial, suggestion: introduce a list of
> compositors / compositor frameworks, as well as clients / client
> frameworks, and which protocols they use and support. This would help
> both application and compositor authors figure out where they should
> invest time and effort. I suggest that we keep this lightweight: have
> a registry of compositors / compositor frameworks / toolkits /
> clients, each with a couple of named people who can speak
> authoritatively for that project.
> 
> We could then allow each project to declare its support (or otherwise)
> for any extension: will not ever implement, implementation not
> planned, no opinion or N/A, implementation planned, implemented but
> use not recommended (or limited/stubbed), implemented and recommended.
> This list would be machine-parseable (XML, JSON, YAML, whatever is
> easiest to fit), with a GitLab CI pipeline used to generate a
> https://wayland.freedesktop.org/protocols/ website on every push,
> which gave both a per-extension and a per-project support table. And
> some more readable docs. I think this would be a really good entry
> point and clear up a lot of confusion.
> 
> As a strawman list of projects to begin with (I'm sure there are others):
>   - compositors and compositor frameworks: Chromium (Exosphere),
> Enlightenment, KWin, Mutter, Smithay, Sway, Weston/libweston, wlroots
>   - toolkits: EFL, GTK, Qt
>   - media: GStreamer, Kodi, VLC, XBMC
>   - other clients: Chromium (client), Firefox, Mesa (EGL/Vulkan)

I too agree that this is a good idea. I'm not too sure about the
usefulness listing all applications that may use some protocol or
another. Might make more sense to keep such a list in a less formal
place like a wiki page on gitlab. For a compositor protocol support
matrix, I see the value as much higher.

> 
> My second suggestion is to formalise the 'xdg' namespace. xdg
> extensions have been accepted or rejected by rough consensus between
> Enlightenment/EFL, GNOME, and KDE. That still seems reasonable enough
> to me, assuming that 'xdg' retains the focus of an integrated (as
> opposed to build-it-yourself) desktop. The IVI namespace would
> similarly be delegated to automotive people, and maybe we could
> delegate the layer_ namespace to those developers as well.
> 
> My third suggestion is to formalise the 'wp' namespace, as core
> extensions that everyone can agree on. It doesn't mean everyone needs
> to implement them, but at least not have active opposition. For
> example, Mutter hadn't implemented wp_viewporter for the longest time,
> but also had no opposition to it being implemented - which wouldn't
> block it being a 'wp' protocol. Or Weston hasn't implemented
> xdg_foreign and probably won't, but I'm fine with it existing and
> being a common extension. On the other hand, Weston/Mutter/etc would
> have very strong opposition to a 'wp_randr' extension, and last I saw
> wlroots would have very strong opposition to wp_pointer_gestures. So
> those wouldn't be wp.
> 
> So where does that leave other extensions? My fourth suggestion is
> that we look to the OpenGL/EGL/Vulkan registries: if an extension has
> been vetoed from the wp_ namespace, but still had support and
> implementations from multiple projects, that we still accept and
> publish it under a different namespace which makes it clear that some
> projects think the extension is fundamentally a bad idea, but it is
> also not a compositor-specific extension. Bikeshedding the prefix to
> use here would be welcome, as I don't have any good suggestions right
> now.

I think formalizing a set of categories, including wp_ and xdg_ and at
least a third one is a good idea. I think that with a "free for all"
method suggested by Pekka, where we put all the responsibility on
everyone else to figure out what protocol fits under what category
(because there will still, in practice, be categories, we just avoid
spelling them out), we loose too much in terms of discoverability and
perceived scope. It'll be harder to know what to expect.

What I mean here is that it should be clear what protocols are
"plumbing" (e.g. wp_viewporter), what protocols are for window
management related tasks where the application aim to be portable in a
sense that they will work on both fully integrated and non-integrated
compositors (e.g. xdg-shell and xdg-decoration), as well what protocols
are for applications that want to be a building block of a
non-integrated compositor (e.g. the layer shell protocol). I don't think
it's possible to avoid politics completely here, no matter how
"annoying" it is, because protocol development is partly about that.
There are most likely good reasons Khronos have chosen to work like
this as well.

A protocol support matrix will still be very useful here either way, and
I expect all categories will include protocols with varying support, as
so is the situation already, but it wouldn't be the only source of
clarity of scope and potential adoption.

We could also keep the door open for other groups of protocols, such as
IVI and maybe even those for fully coupled systems like the content
protection protocols that has been floating around, assuming those who
write protocols agree to act as maintainers and use our versioning
system.

> 
> Once we've established and codified these ground rules - with a
> document along the lines of Wayland's CONTRIBUTING.md - we can open up
> commit access to wayland-protocols, so we're not so reliant on a
> couple of arbitrary gatekeepers. Obviously this would be done with the
> expectation that those ground rules are followed - don't post a wp_
> extension on a Sunday morning and then commit it in the evening with a
> couple of +1s because no-one objected - but we're all reasonable
> adults, and can treat each other reasonably without having to enforce
> ACLs.
> 
> Does anyone have any thoughts or suggestions here? Is this a good
> aspiration? Is it the best way of achieving that aspiration?

There is also the question about what model of contribution and
discussion we should use. Do we rely on merge requests, keeping
discussions there, or do we stay on the mailing list using
git-send-email?

IMHO we should choose one or the other, not some combination where
Gitlab sends E-mails to the mailing list for merge requests, as this
would mean we'd end up with multiple diverging versions of the same
discussion thread.

Jonas

> 
> Cheers,
> Daniel
> _______________________________________________
> wayland-devel mailing list
> wayland-devel at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/wayland-devel