Jerome Kelleher

> Gee, this sounds related to the 'extend edges' thing again! (although maybe only tangentially) Hah, yes! Really great point Peter. This is definitely worth exploring, but we should look...

> So we would still be iterating over the extra segments we create by using ancestral_to. Yes, you're right. So maybe this idea of "keeping more of the ARG" is...

There's a [makefile](https://github.com/tskit-dev/msprime/blob/main/data/Makefile) to download and compile SLiM, so perhaps we should just stick a particular version on that download (something < 4) and make sure that the verification.py will...

Great, thanks. I don't mind which direction we go.

I think this general strategy would work for the DTWF. It would be much more difficult (or indeed not make sense at all as you point out) in the Hudson...

I think this does the trick for a simple three population split model with no migration and different recombination rates: ```python3 import msprime import tskit split_time = 200 demography =...

> It would also be nice to have the nodes relabelled so they conform to the usual time ordering Could run simplify on it, that would rejigger things in the...

(1) seems like the simplest and most consistent with existing models, but whatever makes most sense implementation wise is good by me.

I'm not sure why we didn't document that @daikitag, and I'm honestly not sure whether it's really supported or not. We may have decided it wasn't worth the trouble of...

I'm reluctant to change things like this without a good motivation. Let's have a look tomorrow to see if we can figure out why it's there.