Hello everyone, On Tue, Feb 08, 2022 at 01:06:17PM +0200, Laurent Pinchart wrote: > > My opinion is that if we want to replace something existing with a new > > API or something else, we should first add the new support, block any > > new adopters for the old API such that everyone uses the new API, and > > only after that convert the old API clients to the new API. > > So 'can be reintroduced later on' is not okay. We can't remove things in > > the hope that it would be reintroduced later. Just my personal take on > > this, feel free to have a different opinion. > > When regressions are introduced this makes sense, but here we're > dropping a feature that isn't used as no kernel driver selects a VC > different than 0. I'd like to add that the purpose of virtual channels in CSI-2 is to be able to transport multiple logically separate streams over a single link. That has never been supported in V4L2 so these flags have been unusable to begin with. The streams set a lot of people have been working on on their turn is the way to support virtual channels (as well as other features such as data types), and it does not involve using mbus flags (as the virtual channel isn't bound to the bus). The intent was to cc the patch to Eugen but it seems my gitconfig prevented that. -- Kind regards, Sakari Ailus
