Geoffrey Biggs

We have [a PR in our fork](https://github.com/gbiggs/rosbag2/pull/5) that adds "play until timestamp" functionality. We're planning on upstreaming it later this week or early next week. It might need some revision...

I think that with the rapid ongoing development of a new bag file format, which should not have this problem, the best solution is to fail gracefully and note this...

You can take a look at https://github.com/ros2/ros2_documentation/pull/1663 to see the first tutorial's draft.

That sort of tutorial is on the list to be written. I can't say when it will happen, though.

> Is pre-allocation at process startup sufficient or does it have to be done at compile time? I think this is decided by the sort of microprocessor you want to...

I agree with the need for a general cleanup. Regarding `rmw_destroy_publisher`, what's the reason for the exception? Or does that need changing to not receive an allocator?

I agree with @sloretz that it appears to me to be a perfect example of `ROS_DOMAIN_ID` being useful.

Whether or not to block on publish is something that a bunch of people here and at Honda spent quite a bit of time thinking about a couple of years...

I'm really feeling the limits of GitHub's comment thread management now. > The "best effort + keep all" blocking case was confusing at first, but basically it means "it will...

I can see your point about the impact of allowing external configuration of blocking behaviour. If a node is not explicitly designed for it, then as you say it could...