Status of CASA calibrater tool/tasks
------------------------------------
gmoellen - 06Nov29

o Code refactor underway for >1 year, in earnest the past ~6 months,
  addressing robustness, performance, extensibility, maintenance

o Supported (as of Dec 18) Calibration types:
   (*=new or with new features, #=still devel/testing) 
   - M (Mf)     sampled baseline-based (closure error) solutions
   - B          sampled bandpass
   - BPOLY*     polynomial bandpass (A&P) (time-INdep, from Gildas)
   - G          generic sampled multiplicative complex gain 
   - GSPLINE*   gain splines (A&P) (from Gildas)
  #- J*, (Jf*)  generic polarized gain
  #- D, (Df*)   sampled instrumental polarization (non-linear)
   - P          parallactic angle
   - GAINCURVE  gain(elev) (a la CLCOR)
  #- POINT*     pointing selfcal  (first solvable dir-dep term)
   - T          sampled poln-indep gain (troposphere)
   - TOPAC      global opacity (elevation modulated) (a la CLCOR)

  *New: dual-pol BPOLY,GSPLINE; freq-dep D; J/Jf; POINT

  Anticipated in 2007+:
   - D          apply via *linearized* model (decouple parallel hands)
   - PA         position angle solve/apply
   - TSPLINE    spline version of T solving
   - BPOLY      time-dependence!
   - B,Df,Jf    pre-solve data decimation and smoothing in freq
   - T,TOPAC    via alternative derivations (Tsys data, ATM library)
   - K          fringe-fitting
   - (F         ionosphere, incl dir-dep)

o Supported (as of Dec 18) calibration mechanisms
   - fluxscale (G,T)
   - accumulation (B,G,T) (per type)
   - weight calibration (non-polarized types) (weight zeroing?)
   - interpolation (nearest, linear, aipslin, ?)
   - phase-only  
  #- selectvis w/ MSSelection (no TaQL required!)
  #- smoothing (median,?)   [smoothing is not interpolation]
  #- calplot 
  #- setmodel ("setjy in calibrater")
  #- SNR threshold (!)

  Anticipated in 2007+:
   - rationalize flagging by calibration (flag versions) (weights?)
   - post-solve solution normalization option (e.g., G, B)
   - more smoothing options
   - freq-dep interpolation for applying decimated solutions
   - Amp-only?
   - Robust solving?

o Ambitions:

  - generic scan, field, spw fan-in/fan-out mechanisms, a la
      spwmap, but with additional options (abs/rel freq, etc.)
     -> via accumulate? (a la CLCAL)
     -> automated mappings?  e.g., spw by freq, field by proximity, etc.

  - calibration accumulation model ("SN/CL mechanism")

  - generalized cal-transfer
     -> supported in accumulate, fan-in/out; needs alignment
     -> cf flag transfer