sequence minus needs an automatic cast (possibly with a warning)

in minimize, could pre-process constraints to determine required data
matrices in advance, rather than appending.  Should speed up with many
constraints (although good Opt coding shouldn't require this).

sedumi chokes when # of dual variables exceeds the # of dual constraints -
is this a bug?  It coresponds to a overdetermined primal problem with
no solution and a dual problem with a solution of 0.
Should check for this nonsense case in Opt.

fix mtimes bug in optQuadVector - only allow pos. scaling?
probably need to check optQuad also

> for linear constraints

*equality constraints

*define linear constraints as g{<,==}0, where a is an optVector
so that constraints can be defined before all variables are declared
also requires just one object per constraint
move A & b extraction from constructor to minimize

*define SOC constraints as \|g\|{<,==}h, 
where g & h are optVectors, and length(h)==1
same benefits as above

check all functions that create new types or return classes for
proper propagation of pool info.  i.e. minus() seems broken if
the first term is a constant.

properly overload optVector and then cast all optVector inputs to 
gracefully handle constant vectors.

more fully profile convolution and pwr() functions

fix fftfilt so convolution can use it

*?find bug that occurs when aux vars are used. current workaround is to
change order of variable declaration.

find case dependancies and automate generation of all-lowercase version

check all implicit uses of optVector functions on optSequences, make
sure the type is preserved and we get what we really want.
Including optQuadVector, adding them does return the right thing.

When x is a length-1 optVector, x.^2 (or maybe x^2?) should return a
quadratic, not an optQuadVector?

check for uses of insertion/extraction functions when direct manipulation
is allowed (i.e. we are in a class function of the object)

make sure full complement of operations (+-/*) is available

resolve pwr/power conflict.  power is overloaded ^ function, but is
being used by a lot of programs now.  pwr has no (known) conflicts.

create another level of reporting below warning ('info'?)

add xcorr-like functionality to only computer N lags of corr/conv

display functions

