Antonin RAFFIN

> this order could be useful to code come logic at mask t with the observation at t (even at the t=0 case) This is what is currently done, no?...

I would disagree on the semantic of it. If an environment is already terminated due to some condition, it is not over because of truncation, you should not bootstrap, so...

> We did this as there is no reason why to restrict the truncation (in the general case) to only be True when terminated is False. who is "we"? You...

> The variable truncated means: hint time limit, > reset is optional truncation is not only due to timelimit, see below: > This seems to confuse the meanings of terminated...

> My understanding is that you think that if when a state transitions, it reaches a termination state and the episode ends, therefore never reaches the truncation state making terminated=True...

> With our new API, this limitation no longer exists so I don't think we should limit ourselves just because the old API did. A software with more features doesn't...

This is the current implementation in the compat method yes, but: - I'm arguing it's not intuitive, not needed (you can have the same features with option 1 and info...

let me try to rephrase... > You show below in the documentation that this case of both being true is intended the doc is consistent with the code but there...

@pseudo-rnd-thoughts gym 0.26 has been released for almost two months already and this ~bug~ breaking change is still not mentioned in the migration guide, so it's still time to keep...

> pybullet_envs -> no word on gymv0.26 yet Bullet3 in progress (compatibility with Gym 0.26.0 bulletphysics/bullet3#4335) We should probably patch it the way I did for the RL Zoo: https://github.com/bulletphysics/bullet3/issues/4368#issuecomment-1363269019...