Waltham: Generic Wayland-Style Network IPC

[starshipeleven]

this seems to be a complete alternative to Wayland, with network transparency in mind, albeit implementing a similar API, and apparently a library that would support both backends. I am a bit disappointed, but let's see how this plays out.

This isn't an alternative, this is something that can run Wayland applications over the network with network trasparency (something that Wayland isn't doing and probably will never do), while retaining the same design principles of Wayland itself.

This thing is the base of a remote desktop application that will pwn them all as it will send around data and not rendered frames.

[M@yeulC]

This isn't an alternative, this is something that can run Wayland applications over the network with network trasparency (something that Wayland isn't doing and probably will never do), while retaining the same design principles of Wayland itself.

This thing is the base of a remote desktop application that will pwn them all as it will send around data and not rendered frames.

I am curious, how does it compares to the description which is at the end of my now approved post (just above the one you quoted)? I can't imagine them doing the rendering for every toolkit themselves, and they would have to hook in the program somehow. Unless they are implementing a solution more similar to what I first described, and I got it wrong after re-reading the actual announcement and skimming the documentation?

I also want to add to the said post that passing OpenGL draw command around would be quite impractical, because they would require the transmission of a lot of data; data which is designed with high bandwidth interfaces in mind. So, in this case, we would still probably pass bitmaps around.

[schmidtbag]

I also want to add to the said post that passing OpenGL draw command around would be quite impractical, because they would require the transmission of a lot of data; data which is designed with high bandwidth interfaces in mind. So, in this case, we would still probably pass bitmaps around.

Slightly off topic, but perhaps this would be of interest:

http://www.virtualgl.org/

This allows for GL rendering on remote systems. It's actually a clever way to get 3D acceleration in virtual machines, but unfortunately it's a bit limited in it's capabilities.

[quikee]

I am curious, how does it compares to the description which is at the end of my now approved post (just above the one you quoted)? I can't imagine them doing the rendering for every toolkit themselves, and they would have to hook in the program somehow. Unless they are implementing a solution more similar to what I first described, and I got it wrong after re-reading the actual announcement and skimming the documentation?

I also want to add to the said post that passing OpenGL draw command around would be quite impractical, because they would require the transmission of a lot of data; data which is designed with high bandwidth interfaces in mind. So, in this case, we would still probably pass bitmaps around.

It's all on surface / bitmap level which makes the most sense really. What each client needs to report to the Wayland compositor that a surface changed, maybe it also reports the "damaged" areas.

[starshipeleven]

I am curious, how does it compares to the description which is at the end of my now approved post (just above the one you quoted)? I can't imagine them doing the rendering for every toolkit themselves, and they would have to hook in the program somehow.

I have to admit that the docs and announcement focus on uninteresting descriptions on how this protocol resembles wayland while leaving out basic stuff like "what the hell is this protocol supposed to let me do", clear sign that they have been written by a programmer.

The only parts that matter for our question:
-this thing is a protocol that maps closely to Wayland API but allows to replace network-incompatible parts
-it requires a "software suite" to work (it links to a weston hack that requires stuff on both devices, sending and receiving, might hint that we need something like that also here)
-it has a server and a client component
-it is "very much like Wayland" (+ random technobabble that does not answer the question)


I lend towards something that disguises itself as a Wayland compositor on the server, changes the stuff on the fly to send it over the network and pipes Wayland API to the client where it returns Wayland API so it can be offloaded to a Wayland compositor.

I assume that the application is rendering itself on the server and is sending over the bitmap (I have no data on this, staying conservative), but the compositing/rendering is likely offloaded directly to the client's wayland compositor (as of page 4 of the pdf of the weston hack they linked https://people.collabora.com/~pq/Adi...I-remoting.pdf ).

It SEEMS to do something very similar to what X11 is doing currently, but compatible with any wayland compositor on the client.

[pq__]

This isn't an alternative, this is something that can run Wayland applications over the network with network trasparency (something that Wayland isn't doing and probably will never do), while retaining the same design principles of Wayland itself.

Hi,

it seems you missed this sentence from the announcement: "Waltham is not the magic bullet that makes Wayland network-transparent!"

Waltham is not Wayland-over-network. Waltham is not a replacement, alternative, or an extension to Wayland the display protocol.

[pq__]

I have to admit that the docs and announcement focus on uninteresting descriptions on how this protocol resembles wayland while leaving out basic stuff like "what the hell is this protocol supposed to let me do", clear sign that they have been written by a programmer.

This is because Waltham does not do anything. Waltham, once matured, does not include any protocol for any specific use case. We just can't develop it without any sample XML files.

Waltham is a toolkit for people to easily write their own protocols to do whatever they want to do. It does not have to have anything to do with window systems or graphics.

The only parts that matter for our question:
-this thing is a protocol that maps closely to Wayland API but allows to replace network-incompatible parts

No, the protocol will not be "Waltham".

Yes, we are brewing some kind of protocol, to be carried by Waltham, for some sort of window remoting specific to IVI, but that is all on a layer above Waltham. The stuff above Waltham is the mentioned "software suite". Waltham will implement none of it.

-it requires a "software suite" to work (it links to a weston hack that requires stuff on both devices, sending and receiving, might hint that we need something like that also here)
-it has a server and a client component
-it is "very much like Wayland" (+ random technobabble that does not answer the question)


I lend towards something that disguises itself as a Wayland compositor on the server, changes the stuff on the fly to send it over the network and pipes Wayland API to the client where it returns Wayland API so it can be offloaded to a Wayland compositor.

I assume that the application is rendering itself on the server and is sending over the bitmap (I have no data on this, staying conservative), but the compositing/rendering is likely offloaded directly to the client's wayland compositor (as of page 4 of the pdf of the weston hack they linked https://people.collabora.com/~pq/Adi...I-remoting.pdf ).

It SEEMS to do something very similar to what X11 is doing currently, but compatible with any wayland compositor on the client.

Your description here is very accurate of the final software suite. However, Waltham will implement none of it. Waltham is just a tool that makes it easy to call arbitrary functions over the network.

Everyone here seems to think Waltham is something more than a generic IPC layer. It is not. Waltham is more like RPC (remote procedure call) or a programming language, it has no specific purpose. Everyone can use it for anything they want by writing their own protocol.

[starshipeleven]

Hi,

it seems you missed this sentence from the announcement: "Waltham is not the magic bullet that makes Wayland network-transparent!"

Waltham is not Wayland-over-network. Waltham is not a replacement, alternative, or an extension to Wayland the display protocol.

it seems you missed the part where they say "Waltham could be an implementation detail in a software suite(*) that allows Wayland applications to be ran over a network."

So no, this allows Wayland to be used over the network, but being a protocol it needs some kind of client and server applications to deal with things.

Really, the announcement as-is is horribly confusing.

[pq__]

it seems you missed the part where they say "Waltham could be an implementation detail in a software suite(*) that allows Wayland applications to be ran over a network."

So no, this allows Wayland to be used over the network, but being a protocol it needs some kind of client and server applications to deal with things.

Really, the announcement as-is is horribly confusing.

Yes, I wrote the announcement. I'm the main designer of Waltham. Do you need proof of this?

[starshipeleven]

Everyone here seems to think Waltham is something more than a generic IPC layer. It is not. Waltham is more like RPC (remote procedure call) or a programming language, it has no specific purpose. Everyone can use it for anything they want by writing their own protocol.

As said above, whoever wrote that announcement was unable to explain what the hell Waltham is supposed to do, only thing he did is saying that it can be used to pipe Wayland over the network before spamming random uninteresting technobabble about the technical details about what it is actually made of.

So yeah, people will think it has some connection to that, it's normal.

If you are part of his team, please work together to write a decent announcement v2 where you explain Waltham's scope.