[Spice-devel] Looking for help to start learning Spice through bugfix/design

[Marc-André Lureau]

----- Mensaje original -----
> Thanks for the answer, Marc-Andr é !
> 
> I'll think about these areas of enhancements you've suggested. They need a
> bit of clarification though - because it isn't clear for a newcomer what and
> how to improve :)
> 
> As for OS preferences, I'm a cross-platform guy, though giving a favor to
> Linux naturally. This may be considered as a minus as I'm used to
> cross-platform frameworks which do so much low-level work and hide
> OS-specifics behind nice interfaces... I want to mention that I'm used to
> C++, so getting in mostly C-written components would be harder for me. What
> Spice parts are more C++-friendly actually?

Iirc, the browser plugins (only XPI is public atm), and win-vdagent are C++.

The rest is in C. The Spice client is sharing 99% of the code, while the 
platform-specific bits are handled at lower levels: gdk/cairo etc..
The code is quite portable, thanks to libraries like GLib/GIO/Gtk+.

But there are still platform-specific components that are not portable
(this situation could still be improved).

> BTW, right now I've come across 2 tickets in the tracker:
> https://bugs.freedesktop.org/show_bug.cgi?id=63807( No way to filter
> devices) and https://bugs.freedesktop.org/show_bug.cgi?id=62033 ( Means to
> detect local-only, and related
> https://bugs.freedesktop.org/show_bug.cgi?id=62187 ,
> https://bugs.freedesktop.org/show_bug.cgi?id=62188 ). Are they relevant to
> work on? I need an advice on how to actually fix them (I've added some
> initial suggestions/questions in comments).

There is a lot of improvements possible for local-only, and this is very relevant for desktop applications like Boxes. It is not so relevant for VDI in general, so it's not highest priority. It's indeed a good bug to work on.

Regarding USB filtering, the bug was has a solution and was reopened. Imho, this is quite out of scope of Spice client libraries, so it should be handled at a different level, either libusb* or in the application. Probably in the application until a good API is found.

> 
> For my future work I'm thinking about particular features useful for
> enhancing VoIP clients performance/quality within VDI. One of them is listed
> at the wiki, http://spice-space.org/page/Features/CodecPassthrough . While
> it is useful(and probably should be done at first, or alongside), I think
> there are even better possibilities - when the traffic doesn't go to VM at
> all. How to implement this - it is another question:) One of the most
> obvious ways is to implement something like 'remote media engine' capability
> for spice client, and provide an API to apps running at guest via extending
> vdagent. It should be easier for audio-only solution, for video-capable
> another level of integration would be needed, when this media engine would
> be rendering video in the app-provided window from the guest... Some sort of
> such 'window overlay' APIs are part of Citrix HDX and VMWare Horizon View
> suites actually.

Do applications need to be aware of those API? Perhaps we could implement that? Tbh, I think a Linux solution using GStreamer elements should work. But adapting Windows directshow/media/whatever API to use that framework under the hood might be difficult. We lack of Windows experience to tell what is the best level or API to hook into.

> 
> Suggested simple scheme above requires re-write of media subsystems of
> VoIP/media consumer applications at the guests - not very feasible... Next
> step would be introducing something to support the same for unchanged or
> only very slightly changed applications. I have some ideas on this topic as
> well, though looking hard/unrealistic/unclear (specific virtual network for
> targeted applications, where network traffic is heuristically analyzed and
> audio/video isn't transferred but used to control clients' media; signaling
> traffic pass-through).
> 
> All this seems to be very exciting topic to me, with many different
> approaches and directions actually possible. I'd like to hear your opinions
> about it, discuss it.