Waypipe Is Successfully Working For This Network-Transparent Wayland Apps/Games Proxy

[kpedersen]

lol, i've found someone not aware of youtube in 2019

It goes without saying that Youtube works because it can buffer. A "live" stream from a desktop at a high resolution is unable to buffer.
Either disable buffering on Youtube or watch a live screen at maximum "quality" and you will see the issue.
Also, if youtube needed to do the rendering on its servers (like you will have to with Wayland), it would quickly become unsustainable.
Frankly the fact that you brought up youtube suggests you don't quite understand the general issue at hand when it comes to remote rendering a desktop session.

this solution is already smarter than your imagination

It looks like all Wayland users might be having to use their imaginations for remote computing if they are unable to solve this XD

How is Citrix doing it in comparison?

The RDP protocol was originally developed by Citrix in partnership with Microsoft (check out Citrix WinFrame for a quirky Windows 3.1 style terminal services experience). Even to this day, their solution is largely similar on Windows (you can even still connect to the old WinFrame OS using their latest viewer) and on Linux it relies on X11 (like NoMachine/NX, it does some clever prediction / compression)

[starshipeleven]

What a shame.
....
Hopefully a smart and modern solution will come in the future

Wayland's whole concept is about having application render their stuff on their own, and then send the rendered frames to the compositor. There isn't much choice, you are working with rendered frames.

Why are you still expecting it to do X11-like remote-rendering (for applications that actually support that), it's against the design principles.

[starshipeleven]

It goes without saying that Youtube works because it can buffer. A "live" stream from a desktop at a high resolution is unable to buffer.
Either disable buffering on Youtube or watch a live screen at maximum "quality" and you will see the issue.

Sorry what? Buffering is not relevant, people don't launch a video and then go brew a coffee and come back later to watch it. That's people torrenting, you ancient dinosaur (and even then there are torrent-based streaming applications too for people with fast enough Internet).

People on the Internet launch a video and expect to view it RIGHT NOW (insert jokes about spoiled young people). Buffering is happening, but download speed is always faster than view speed. In many cases the Youtube player stops buffering too to avoid wasting bandwith downloading the whole video (in case the user stops and closes the page).

Youtube (and Netflix) work because they lower the bitrate and do other compression tricks (and adapt the resolution/quality on the fly). The stream is NOT as large as the original video.

https://www.tutorialguidacomefare.co...esses-youtube/

[tildearrow]

So can this thing do AV1?

Wow, only 1 month has passed and you already ask for AV1.

Maybe, if you can tolerate a super-low frame-rate (which isn't the goal of this project), sure, wait a year or so, and it will use the AV1 video encoder.

[kpedersen]

Sorry what? Buffering is not relevant, people don't launch a video and then go brew a coffee and come back later to watch it.

You're joking right? Why does Youtube bother buffering then? They still need to do it due to latency. You might have a really large bandwidth (so can download fast) but if you have large latency / ping, live footage will give a poor experience... So you buffer videos, again which you cant do with a live stream.

Why are you still expecting it to do X11-like remote-rendering (for applications that actually support that), it's against the design principles.

Its beyond Wayland's scope which is why everyone is going to keep having to use X11 on top (for remote work); so nothing has really been achieved.

[uid313]

Wow, only 1 month has passed and you already ask for AV1.

Maybe, if you can tolerate a super-low frame-rate (which isn't the goal of this project), sure, wait a year or so, and it will use the AV1 video encoder.

Hmm, okay. So what's good then?
If AV1 is bad. Is VP9 better or H.265? Or maybe some special purpose low-latency video codec?

[starshipeleven]

You're joking right? Why does Youtube bother buffering then? They still need to do it due to latency.

Can you please explain in detail how latency matters for a video stream that does not care about input, and how buffering can affect latency in your alternate reality?

In this world when you look at a video you will be seeing stuff X time later after it was sent from the download servers, and that's it, that's the "latency" for a youtube video. This happens regardless of how good or bad you buffer. The minimum latency is the travel time of the media stream.

Buffering (again, in this world) happens for the same reason the old CD players also buffered the music they were reading. Internet speed isn't constant nor guaranteed. To ensure a SMOOTH playback of the media even if there are drops or temporary hiccups in the download, you buffer it.

As I said, videos are meant to be played at a SLOWER rate than their download speed so they can play smoothly until the end, so buffering isn't used as a "fix" for the low bandwith

You might have a really large bandwidth (so can download fast) but if you have large latency / ping, live footage will give a poor experience... So you buffer videos, again which you cant do with a live stream.

No, no no. Streaming does not work like that. Streaming is bulk data transfer over UDP, a protocol that does not even check that you got the data. Latency for a video stream is perceived only as the delay between the time you press "play" and the moment video starts playing. After that, even if you are seeing it "not truly live", but with some delay, your viewing experience won't be harmed.
This happens all the time with TV broadcast, that is several fucking seconds late (huge latency).

What can screw over a video stream is inconsistent bandwith, i.e. you don't have enough bandwith to send over data fast enough to show, OR shitty connection, i.e. you lose data along the way.

Latency matters when there is input involved, like gaming or remote control.

Its beyond Wayland's scope which is why everyone is going to keep having to use X11 on top (for remote work)

Not a lot of people uses X11 for "remote work" like in the good old days, and This has more to do with X11 being an obsolete piece of shit than anything.

Everyone uses some kind of application that basically takes screenshots or videos and streams them (Teamviewer, Anydesk, NoMachine, VNC, RDP even if theoretically since ver7 should be able to do better).

Because it's apparently better than relying on the aging X11 to send over stuff, requiring both less bandwith and also much less latency (as X11 remote requires many round trips to actually render stuff on the other machine, with a media stream you are sending the images ONCE). https://superuser.com/questions/1217...so-inefficient

[tildearrow]

Hmm, okay. So what's good then?
If AV1 is bad. Is VP9 better or H.265? Or maybe some special purpose low-latency video codec?

H.264. We have x264 (very fast) and it's widely supported by hardware.

AV1 isn't bad, it's just that the current implementations are slow.

[smitty3268]

Its beyond Wayland's scope which is why everyone is going to keep having to use X11 on top (for remote work); so nothing has really been achieved.

Are you under the impression that X11 does something different than just dumbly passing rendered buffers around? Because it doesn't. This is essentially the same thing, only smarter because Wayland has the concept of frames.

[smitty3268]

Hmm, okay. So what's good then?
If AV1 is bad. Is VP9 better or H.265? Or maybe some special purpose low-latency video codec?

AV1 will be interesting once hardware acceleration is available. Until then, it's way too slow to use for anything realtime like this.