DONE

* Add covariance statistics functions
   -- Still need to think about weighted version of covariance functions

* Make the algorithms work with rectangular MxN matrices where
appropriate. This is very confused right now.

* Clean up the error handling, _impl is not necessarily the way to go
where there are potentially fatal errors.

* LU solver seems a bit unstable on the Vandermonde problem. More
precisely there is a big difference when optimization is switched on
(surprizingly optimization makes the accuracy much worse).

  WITHOUT OPTIMIZATION            |     WITH OPTIMIZATION (-O4)
                                  |
 12[0]: -3.14366197229564272e-18  |   12[0]: -2.66822535639952236e-13
 12[1]: -7.70076451716862649e-18  |   12[1]: 1.81545260902433614e-11
 12[2]: 5.92048153457652747e-16   |   12[2]: -5.44438612354179128e-10
 12[3]: -5.3792458564953351e-16   |   12[3]: 9.47672001543103068e-09
 12[4]: -3.33857096865221473e-16  |   12[4]: -1.0607961716333582e-07
 12[5]: 2.77580003807683497e-16   |   12[5]: 7.99037652257140137e-07
 12[6]: 7.73147382223194456e-16   |   12[6]: -4.11543937371223868e-06
 12[7]: -1.08294074021194352e-15  |   12[7]: 1.44182659189005111e-05
 12[8]: 8.8462083488842609e-16    |   12[8]: -3.34497439296624311e-05
 12[9]: -7.68742511878438236e-16  |   12[9]: 4.85121346376069841e-05
 12[11]: -1.03963268321917673e-16 |   12[11]:  1.304008127082465e-05   <****
FAIL:   solve_LU vander(12)       |  FAIL:   solve_LU vander(12)

* Subtle bug(?) in BLAS. If uninitialized memory is passed with a
multiplier of zero then it is accessed and multiplied by zero. This
would be ok, except that the memory could contain NaN or Inf. Should
check for zero multiplier and explicitly handle it.
  
  Don't pass uninitialized memory!  You might want to have infinities
  and NANs in a vector and special casing the multiplier == 0 would 
  lose the correct IEEE properties for these quantities

* Upgrade to texinfo-4.0.  

* Definitions of macros which are negative numbers should use
parentheses, e.g. in gsl_machine.h we need #define GSL_LOG_DBL_EPSILON
(-3.6043653389117154e+01) instead of -3.604...

* Fix texinfo files so that sentences consistently use a double space
after the preceding period (n.b. do this by hand, because there are
names in the references which need only a single space,
e.g. D. E. Knuth).

* In matrix struct change dim2 to tda, otherwise the distinction of
size2 vs dim2 is totally confusing.

* Should doc/mdate-sh be in CVS? Automake needs it unless you supply --missing

No. The automake command in autogen.sh has --missing.