Removing myself from reviewers. I had a couple of minor comments on the obs_base and obs_lsst PRs, and of course the big one is export/import migration in daf_butler - but I think we should probably (mostly) punt on that on this ticket, and just limit ourselves to some combination of:

• handling just a few special cases in a not-so-principled way to avoid breakage for CI-tested git-committed export files;
• just putting the right versions into export files so we can recognize problems going forward.

Sorry that took me a little while to get to. I can't remember all the things that have been mentioned for the dimensions.yaml part of the review, but I do trust that they were captured on the various tickets, so as long as they're all in there, I'm confident this should be everything needed. I left a couple of minor comments on the translation code.

Actually, one of the big things that is very much needed is DM-30948, and which I thought was coming through in this schema migration cycle, but looks like it's not.

The lt;dr there is that visitIds are only derivable via an awful hack, and data which goes in as an exposure and comes out as a visit cannot be retrieved expect via this hack, and that's a very bad look for users.

Merlin Fisher-Levine: The entire purpose of this schema change is to implement DM-30948 (see the first line of the description of this ticket). The code you reviewed on obs_lsst calculates the quantities that are needed and if you look at the obs_base PR you will see that it is all devoted to getting the new visit definition working.

Please look at the dimensions.yaml changes on the daf_butler PR which implement all the record changes you requested.

OK, great. I couldn't remember/didn't know exactly what was required to do that, and just saw that that ticket was still To Do and figured I should mention, but glad it's all covered. (I had read all the bits you said, like I said, I just wasn't conscious of what was required for that other ticket)