quasar987 said:
Another one of the physics teacher working at the college I attended (see my
previous post) wrote a book on basic http://universite.deboeck.com/Resources/Titles/28011100266200/Images/28011100266200L.gif . In this book it says that the expansion of the Universe is due to the expansion of space itself and that this expansion of space is entirely predicted by Einstein's theory of GR. Now, if this is so, where does the DARK ENERGY (suposedly a MYSTICAL force of repulsion between matter) fit it ?!
This question is about theory; at least as important - since this is science, arguably more important - is experiment and observation.
AFAIK, there are two classes of observations (no experiments

) which support 'dark energy', distant supernovae and 'all cosmologically relevant observations'
1) Distant supernovae. Ever since Hubble first published obsevational results which pointed to an expanding universe, astronomers have been keen to characterise that expansion as accurately as possible - over all distance scales, and in all directions. To do this, they need independent measures of both distance and recession speed. The latter is relatively easy to measure, with great precision - the 'redshift' of galaxies and quasars; the former has proven immensely difficult, and only in the last decade has a broad consensus emerged (there is still considerable 'observational error', and there are still some dissenters). Distance can be measured by a number of different types of observations, and the 'distance ladder' is now reasonably well established, out to perhaps as much as 10 billion light years.
While redshifts are relatively easy to measure, their interpretation as 'recession due to the expanding universe' has had its challenges and upsets. Perhaps the biggest was 'http://cow.physics.wisc.edu/~ogelman/guide/gr8a/ '; apparently there is an awful lot more mass in the local part of the universe (~100 Mpc) than can be 'seen'.
So, the 'expanding universe' hypothesis predicts that the universe is, and has been, expanding uniformly - in all directions, and at all times (at least, after any inflation ended; observationally we can't directly see any earlier than ~300,000 years after the BB, well after any inflation finished), over sufficiently large chunks of the universe (superclusters have sufficient mass so 'expansion' of objects within a supercluster may be small compared with net gravitational attraction from masses within the supercluster).
What do observations show? The best distance indicator for very distant objects is Type 1a supernovae - we think we understand their behaviour sufficiently well so good observations of their lightcurves can be turned into estimates of distance (
this 1998 poster gives more details). Plotting distance against redshift shows that the data are not on the curve predicted by 'uniform expansion' (it's not quite as clear-cut as this; different models of the universe - e.g. with different amounts of dark matter - give different curves, but no model that is consistent with other observational data - e.g. WMAP - goes near the data). What model curves will go through the data? Those in which the rate of expansion of the universe is increasing!
'Dark energy' is the shorthand that is used for whatever might be causing the observed acceleration of the universal expansion.
But not so fast! How can astronomers be so certain that there aren't other effects involved? Initially, quite a bit of work had to be done to nail these down, e.g. do we understand Type 1a SN sufficiently well? what if the distant SN are partly obscurred by dust? and so on. Many of these 'systematic errors' have now been characterised and their effects on the data taken into account. However, some feel that at least an OOM more of good data are needed (
http://snap.lbl.gov/ to the rescue?)
2. All 'cosmologically relevant observations': As Garth correctly points out, this is a bit of chicken and egg. Basically, you take all observations that have relevance to cosmology - WMAP and others on the CMBR, large scale structure (e.g. 2dF, SDSS), primordial nuclide abundances, the Hubble relationship, the distant SN data, ... - and see what sort of cosmological model is most consistent with it all. The idea is that there is more than enough data to constrain models, so if the models are wrong, there will be obvious inconsistencies. The good news is that there are models which are consistent with all the data ... but only those which have 'dark energy' in them!
Further reading:
Wikipedia
APOD
Snowmass 2001, Yellow Book on Dark Energy
short article - model-independent dark energy
nice, bite-sized physicsweb summary