How much should this be compatible with Yasuda-san's survey ops db? (See the schema pages 7+)

The obvious overlap is the Exposure table. I think that the observing schema needs to be split, with the core part merged into AIT's exp and the observing conditions and fiber allocation details pulled out. Something like:

CREATE TABLE Exposure
(
  --    include Pxyz and seqno, to allow merging 
  --    from multiple sites.
  exposureId CHAR[10] PRIMARY KEY, 

  --    denormalize frameId for convenience. Or
  --    add functions.
  visitId INTEGER NOT NULL,
  site CHAR[1] NOT NULL,

  --    add WITHOUT TIME ZONE?
  obsdate TIMESTAMP,
  exptime FLOAT,

  --    bias, dark, arc, flat, object, junk
  exptype TEXT,

  --    OK, bad
  quality TEXT DEFAULT 'OK',
)

We also need to know which cameras are part of this Exposure. Note
that the individual cameras will have their own exptime and obsdate, but we do not track that here. We could. And the NIR 5.57s*N exptimes argue that we have to.

CREATE TABLE CamExposure
(
  camExposureId PRIMARY KEY,

  exposureId CHAR[10] NOT NULL,
  spectrograph INTEGER,
  arm CHAR[1], 
)

Yes, there will be up to 12 CamExposure rows per Exposure.

Comment from fmadec copied from INSTRM-114:

The current very simple exposure logs of LAM is there:
https://people.lam.fr/madec.fabrice/pfs/

I've almost covered everything fixing tickets/INSTRM-274. The mechanism to fill experiment table is not working yet .

The scheme is very similar to Craig's proposal

1. Create table
   c.execute('''CREATE TABLE Exposure
                (exposureId CHAR[10] PRIMARY KEY, 
                 site CHAR[1] NOT NULL, 
                 visit INTEGER NOT NULL,
                 obsdate TIMESTAMP,
                 exptime FLOAT,
                 exptype TEXT,
                 quality TEXT)''')

1. c.execute('''CREATE TABLE CamExposure
               (camExposureId CHAR[12] NOT NULL PRIMARY KEY,
                exposureId CHAR[10] NOT NULL,
                spectrograph INTEGER,
                arm CHAR[1])''')

1. c.execute('''CREATE TABLE Experiment
                (experimentId INTEGER PRIMARY KEY,
                 name TEXT,
                 visitStart INTEGER NOT NULL,
                 visitEnd INTEGER NOT NULL,
                 type Text,
                 cmdStr TEXT,
                 comments TEXT,
                 anomalies TEXT)''')

For now the quality flag is by default set to "OK". 
But we have to be careful because the definition of "OK" is certainly not the same for us, rhl or ncaplar.

merged in 428f8d0