MAIN CHANGES (w.r.t. version 3.024)
* Speed: about 2* faster. Now the speed is comparable to the TI92+,
Erable is sometimes slower, and sometimes faster: most of the time, the TI92+
is faster for small problems (where the processor speed is important)
and Erable is faster for complex problems (where efficient algorithms
are important)

* New functions:
PROMPTSTO: example 'A' PROMPTSTO: asks the user for the new value of 'A'
FOURIER: 2*pi periodic function, integer n -> exponential n-th Fourier coeff
ATAN2S, ASIN2T, ASIN2C, ACOS2S: inverse trigonometric functions conversion
SOLVEX: solve for the current variable
ZEROS: like SOLVE, but without 'x='
TAYLOR0: Taylor expansion at x=0 with respect to the current var
(relative order 4)
PLOTSTK: plot stack level 1 (erase graph, stk1 -> EQ calls PLOT inform)
PLOTADD: add level 1 to plot (add stk level 1 to EQ list and calls PLOT inform)
Both PLOTSTK and PLOTADD return their argument, hence may be in an algebraic.
STOSYM: put level 2 in name level 1 and returns level 1, can be used in an
algebraic.
KEYEVAL: execute key code (like user-key assignements). If arg is negative,
execute the standard key assignment. Example: -63.3 KEYEVAL -> PLOT inform.
MENUXY: return the list of Erable commands whose lid are between stk2 and stk1.
XGROB obj -> grob, tries first EQSTK/JAVA AGROB command, then MK ->GROB2 then
built-in ->GROB2
GROBADD: grob1, grob2 -> grob1+grob2 (horizontal gluing).
SCROLL: grob -> (display grob)
TABVAL: symb, liste -> liste (evaluates symb with cur var=element of lists)
MAIN: a menu-interface to Erable (27SF to get a choose-menu)
ERCFG: make real/cmplex, symbolic/numeric and integer arit/polyn. arit
mode changes easier.

* New functions in add-on libraries
POTENCE: step by step euclidean division
PFSTEP: step by step partial fraction expansion
CHINREM: Chinese remainder. Stk2 { n N } Stk1: { m M }
returns { o O } such that o mod N=n and o mod M=m
...MOD: modular operations, including EXPAMOD for EXPAND and COLCMOD for
COLLECT (currently works only for mod-square free arguments)
P...: internal Erable operations, minimal check.
Ex: 'X+5' { X } S2L 10 PPOWER SWAP L2S
TABVAR: variation table of a rational function
LAGRANGE: Lagrange polynomial interpolation.
{ { x1 ... xn } { y1 ... yn } } -> P s.t. P(x_i)=y_i

* Changes for existing functions
SOLVE: is now multivariate.
Example { 'X^2+Y^2=1' 'X+Y=1' } { X Y } SOLVE
LDEC: returns only the general solution on stack level 1 using Y0, Y1, ...
for y(0), y'(0), ... (scalar case) and V1, ..., Vn for v(0) (1st order
diff. system)

By version:
3.124: SETUP program added for the fixed RAM address versions.
3.121: fixed RAM address versions, see absolu.inf
3.117: inverse trig. functions conversions. Removed CONV from menu.
3.116: IPP accepts now a syntax adapted to integrals without bounds:
stk2: u'*v, stk1: u -> stk2: u*v, stk1: -u*v'
(The current variable is used for derivation. Warning: IPP assumes that
u'*v does not contain any integral)
Example: 'ASIN(X)^2' X IPP 'sqrt(1-X^2)' IPP EXPAND INTVX + + returns
the antiderivative of 'ASIN(X)^2'
3.109: TCOLLECT improved
3.108: fixed real/complex mode in SOLVE
3.106: fixed a bug in fraction derivative and TRIG

3.104:
Changed a lot of names to stick as far as possible to TI92 function names.
This might look strange to HP users, but my objective is to make it easier
for users (especially teachers) to go from TI to HP calcs. HP users are now
much more isolated in the education world than Mac users in the PC world, we
must improve compatibility.
EXPA -> EXPAND
COLC -> FACTOR
DSOLVE -> DESOLVE
LNCOLC -> LNCOLLECT
PF -> PARTFRAC
FACTO -> FACTORS
idn -> IDENTITY
TEXPA -> TEXPAND 
EXPLIN -> LIN
TRIGLIN -> TLIN
TRIGCOLC -> TCOLLECT
INVL -> inv
XY -> DOTP
WEDGE -> CROSSP
FSIGN -> SIGNE

3.102:
* Reintroduction of a SX version (I have rewritten the GNU parts of the solver)
* Fixed a bug 'sqrt(x)=y' x EXEC

3.101: first official beta release
* Function name changes: 
RISCH -> INTVX
der1  -> DERVX
SOLV  -> SOLVE
* Row reduction: two algorithms are provided. If you set flag 20,
the Bareiss algorithm is used instead of GCD simplification, you can get 
the exact inverse of a 20x20 matrix in 544s (faster than the TI92+!).
For small matrices however, this algorithm is slower, hence flag 20 is
cleared by default.
* User keys: GXKEYS works now for Metakernel users too. GXKEYS provides
only the Erable menu-interface. The main menu is available using UKEYS.
* Removed automatic installation of MinWriter from SETUP utility to have 
more memory free (about 25K free which makes the 48G+ like the 48G but with
Erable installed)
* Some programs were moved to 4 add-on libraries.

Future possible enhancements for version 4 of erable:
* Improved ode solving
* Inequalities
* Sequences and series (discrete antiderivatives) 
* Berlekamp algorithm on Z[i]/pZ for p prime such that p mod 4 = 3,
(-> full factorization of polynomials over Z[i]),
* long precision float

