The simplicity of the hybridization technique is its main advantage. It is
simple to understand and simple to implement. However, this method works
badly in practice because because it is truly difficult to obtain a
reliable deconvolution of interferometric data alone when short-spacing
information is important. Indeed, a multiplicative interferometer filters
in particular the zero spacing. This implies that the total flux in the
dirty image is zero (*i.e.* as much negative as positive flux in the dirty
image) but that the dirty beam integral is also zero (*i.e.* as much negative
as positive sidelobes). When we add the short-spacing information (and in
particular the zero spacing) through the pseudo-visibility method, we
enforce positivity of the dirty image total flux and of the dirty beam
integral. It is well-known that trying to deconvolve a mosaic built only
with interferometric data is quite difficult. It almost always requires the
definition of support where the `CLEAN` algorithms can search for clean
components with the clear risk to bias the final result. In contrast,
adding the short-spacing information through pseudo-visibilities enables an
almost straightforward `CLEAN` deconvolution *without* the need of
any support.

For the sake of illustration, let's assume an intensity distribution made
of a large scale structure (*e.g.* a uniform intensity) superimposed with a
small scale distribution both in emission and absorption. A multiplicative
interferometer will filter out the uniform intensity distribution. If there
is no additional zero spacing information, the uniform intensity
distribution is completely lost with the important consequence that the
final deconvoled image will have positive (emission) and negative
(absorption) structures. Trying to reproduce both negative and positive
structures is one of the most difficult task for deconvolution algorithms.
Algorithms of the MEM family enforce positivity in the deconvolved image.
In addition, the presence of large negative structures makes instable the
algorithms of the `CLEAN` family (because it is difficult to distinguish
between negative absorption structures and negative sidelobes of emission
structures). Only the definition of support around positive emission peaks
succeed to stabilize the `CLEAN` algorithms with the drawback of biasing
the result.

Both kind of algorithms (hybridization and pseudo-visibility) are
implemented in GILDAS. However, we strongly recommend to use the
pseudo-visibility algorithm. That's why only the pseudo-visibility method
is packaged in a user-friendly way (*e.g.* through the `Short Space
Processing` widget). (,,) showed through
simulations that 1) the pseudo-visibility algorithm implemented in
GILDAS enable extremely reliable results (fidelities of a few thousands)
on ideal observations and 2) the accuracy of the wide-field imaging is
limited by pointing errors, amplitude calibration errors and atmospheric
phase noise (and not by the used algorithms), even for ALMA.