[pzo]

> Also frankly, nobody forking Flutter will be nearly skilled enough to work on the Flutter engine (Impeller).

For me its so weird they ditched google/skia to develop Impeller in Dart from scratch in the first place. If skia was not ready they could move just 1-2 developers to Skia team to collaborate with them. Now they want to even write their own 3d rendering lib based on Impeller (flutter_gpu, flutter_scene) even thought google has already mature 3d engine in C++ (google/filament). For me from the outside looks like "not envented in our team" syndrome. We are now in funny situation that react native has react-native-skia and react-native-filament and Flutter teams reinvents the wheels.

[xster]

https://github.com/flutter/engine/blob/main/impeller/docs/fa... discusses the reasoning a bit. There's more of a difference in the approach

[pzo]

Related https://groups.google.com/g/skia-discuss/c/Fzpne-f9gZ4/m/5_I...

Seems like most issues could be solved with new Skia graphite backend

[Vt71fcAqt7]

Also related: https://github.com/flutter/flutter/issues/77412 - Hixie's issue that led to Impeller (or one of many issues)

Some related iOS jank issues from that time: https://github.com/flutter/flutter/issues/32170 https://github.com/flutter/flutter/issues/61450

[nmfisher]

As someone who has slowly been writing Dart/Flutter bindings for Filament[0] over the last couple of years, I agree the internal 3D engine was a bit of a strange choice of priorities.

[0] https://github.com/nmfisher/thermion

[irq-1]

The build process for skia and its dependencies is huge. Flutter compiles (on windows) with CMake and Skia uses googles GN tool.

https://skia.org/docs/user/build/

https://gn.googlesource.com/gn

[pzo]

They been building it before and still building and releasing with flutter so it's not like they don't have pipeline for that. And it was distributed compiled so nothing to compile for 3rd party users.