Q1 - Why the image map of the object is computed from the (u,v) map
     (visibilities) by using an FFT ?

  Indeed, we could get a map of the model by computing directly its
  brightness on a grid (x,y). Indeed, we could compute directly disks,
  gaussians, ellipses, etc. on this grid. But if the model uses Dirac
  functions, it is quite unlikely that the grid will contain the exact
  coordinates of the Dirac functions ! So we will miss the Dirac
  functions. Actually, the map must be understood as a "density" function.
  We need somehow to interpolate the model. Going through the (u,v) map
  overcome the problem. This method will also make the maximum resolution
  of the map consistent with the chosen pixel size.

  Also, asking the user to be able to compute both the visibilities and
  the brigthness would twice the work to be done and could introduce bugs
  and artefacts (visibilities and brigthness must have the correct
  relationship).


Q2 - In the model input file, why the values of the parameters are defined
     separately from the functions, outside target field ?

  Because you can use the same parameter (the same name) in different
  functions or event, for different targets. Values, bounds, etc must be
  defined only once.


Q3 - What is the meaning of "scale" for parameters ?

  It is a great help for the fitter to know the order of magnitude of a
  parameter. Indeed, we want to minimize quite non-linear critieria with
  parameters of different meaning (intensity, position, temperature) or
  units. For example, the fitter has no way to know if you are using
  milliarcsecs, arcsecs, or radians for a given parameter. So what can be
  the meaning of doing a step of 1 in a given direction in the parameter
  space ?

  The scale means the order of magnitude of the parameter. The fitter will
  actually use the value of this parameter, normalized by scale. If not
  given, an attempt to estimate a scale will be used. After scaling, all
  the parameters should evolved in the range [-1,1] approximately.