Some concrete changes to remember to do (mostly from Russell Owen and Pim Schellart [X]):

• Drop UNIQUE constraint from Sensor.name
• Section 2.3: “nnames” -> “names”, "None null" -> "NULL"
• page 6: “...potentially entries one or more metadata…” insert “in” before “one or more”
• Execution table description "A unique autoincrement field used as part of the primary key for Execution.” (it is the primary key, not part of it)
• “Run is specialization of Execution” -> “Run is a specialization of Execution”.

More general questions:

• Do we need the Datasets a Run points at to have DataUnits or be a part of Collections? What about Configs (or are those in the pipeline_id Dataset?)
• Need to make sure all angle fields specify units.

Units and reference frames.

Should add ra, dec to Tract table.

Changes in daf_butler, skymap (trivial), and dmtn-073.  I'm not sure the Python code in the technote really needs to be reviewed, but of course you're welcome to take a look (though looking at the diff with what used to be there is probably even less useful).

Looks fine. I discovered a daf_base dependency that I hadn't noticed before that concerns me.