Last major update: 20/06/06

20/06/05 * Added all data for 60000-69999.

09/06/05 * Added all data for 50000-59999.
	 * renamed files e.g. curves.00000-09999 etc;  no changes to
           content since no elliptic curves have conductor divisible by 1000.

27/05/05 * Added all data for 40001-50000.

05/05/05 * corrected generators for 14 curves in range 30001-40000
           which were not saturated, with consequent corrections to
           allgens.30001-40000, allbsd.30001-40000,
           allbigsha.30001-40000 and shas table.  (None of these were
           optimal curves so gens.* unchanged)

03/5/05 * corrections to files allgens.30001-40000 gens.20001-30000
			       allgens.20001-30000 gens.30001-40000
			       allbsd.00001-10000 allgens.00001-10000
	  thanks to Geoff Bailey.  (Some duplications, and some wrong
			       numbering of curves in an isogeny class).

26/4/05	  * added curve 25350III2 which had been missed

22/4/05:  * added data for range 30001-40000 (thanks to Nottingham's
         	  multiprocessor GRID)
	  * resorted files into uniform ranges of 10000
	  * only include gzip-ed files

 1/3/05:  added data for N=15299 (one curve) previously omitted

 9/2/05:  added data for range 25001-30000 + certification of optimal
	  curves extended from 8000 to 11000 (extended to 12000 on 13/02/05). 
	  Swapped the two curves in classes 15180,15624,15744 as the
	  conditionally optimal one was second not first.

21/6/04:  added data for range 20001-25000 + certification of generators
	  for all curves so far

 9/3/04:  corrections for 15810U3 -- previous generator was
                                     3*generator;  so |Sha|=9 not 1.

 4/4/03:  added data for range 17001-20000.
12/2/03:  added data for range 16001-17000.
17/1/03:  added data for range 15001-16000.

25/10/02:  corrected torsion order for 4830N4 in allcurves.1-8000
	   mention 2 new curves of rank 3 in INDEX file

8/10/02: Added data for the range 12001-15000, with the same status as
for the range 8001-12000 detailed below (except that Mark Watkins has
not -- yet -- covered this range). [He has now: 25/10/02.]

14/1/02: Curves in classes 8160R, 8585C, 11024B reordered.  For these
levels the first curve in each class is guaranteed Gamma_0(N)-optimal.
For other levels up to 12000 where there is no guarantee, our first
curve is the one which has been /conditionally/ proved optimal by Mark
Watkins.  It is now the case that for all levels in the range
8001-12000 the first curve in each class is either proved optimal, or
at least agrees with Mark Watkins's conditional list of optimal
curves.  The former holds whenever the degree of the modular
parametrization is given in the curves file.

12/12/01: for around 10% of classes in the range 8201-9000 the
ordering has been changed.  In the new order the first curve in each
class is more likely (though not guaranteed) to be optimal.  This
change makes this range compatible with the range 9001-12000 in that
the same strategy is used throughout for finding a curve from each
newform.

November 2001:

The tables are extended to level (conductor) 12000.

   For levels up to 8000 fuller data exists, including determination
   of the optimal curve in each isogeny class and the degree of the
   modular parametrization.  The numbering of the curves in each class
   (with the optimal curve #1, except for class 990H) may be taken as
   canonical and standard in this range.

   For levels 8001-12000 we give all curves in each isogeny class, but
   the first curve is not known to be optimal (unless the class only
   contains one curve, of course).  The numbering of the curves in
   each class containing more than one curve should NOT be taken as
   canonical and standard: it will change if the optimal curve turns
   out to be not the currently labelled first curve.  It is likely
   (based on past observation) that, where the first curve in a class
   is not optimal, it is related to the optimal curve by a 2-isogeny.

   In the range 8001-12000 we also do not have the modular degree (in
   most cases).

   For any individual level in the range 8001-12000, it is possible to
   carry out the extra computations needed to fill these gaps
   (determining the optimal curve and modular degree), but doing so
   for all curves in the range would probably take a few months with
   the current algorithms and programs.  An independent computation by
   Mark Watkins <watkins@math.psu.edu> has determined, conditionally
   on the Stevens conjecture, which curve in each class is optimal,
   and has determined its modular degree.  Our results agree for all
   levels up to 8000 and we expect them to agree in the higher range
   also.

17/9/01: Error corrected: added 5104B1 to files, omitted due to an
error discovered while working on 2*5104=10208.  Further checking has
shown that no further omissions were made at around the same time.

(for previous update messages see the files sept2001 etc.)