1.3.1   Fixed bad cut&paste memory problem in SetSize
1.3     Fixed problems with stack temporaries being passed as non-const
            references by making *= etc. in-class for Vec[fd].  This is a
            temporary workaround. It breaks some sub vector stuff, such as
            sub(v, 2, 4) *= 3.0; The latest version of gcc shipped with
            redhat 7 no longer takes the -fpermissive flag, however, so it's
            necessary. A redesign to get around stack temporaries no longer
            playing well with constness (and thus necessitating separate
            const and non-const helper classes, and making it harder than
            ever to get C++'s one-conversion-only type system to do what it
            needs to do) is in the works -- it will be VL version 2.0.
        added SubVec -=/+=/etc.
        eliminated v *= m;
        removed all asserts checking that the return value of 'new' is
            not zero. The C++ 'standard' has been changed so that new
            instead throws an exception in this situation.
        added option to use memcpy for general vector copies: VL_USE_MEMCPY.
            The situation seems to have changed so that, on Intel processors
            at least, using memcpy is faster than loops for most cases. See
            note in Vec.cc. (ADVANCED: If you want a container type that
            needs to have its constructor called, this should be disabled.)

1.2     Solve/factor stuff removed from VLf
            who would ever want heavy duty numerical routines in float?
        changed headers to assume vl/VLd.h, vl/Vec2.h etc
        set VLConfig.h on install, so programs linking against it get right
            flags and don't have to do -D__GCC__ etc. (duh.)
        added Normalise() method to vectors... normalise() is now deprecated.
            (The lower-case vector ops should only be math functions.)
        added zero-length assert to norm and Normalise

1.1.4   minor VLMath.h fixes
        fixed buggy sparse matrix trans()
        added ConvertVec and ConvertMat for converting between float/double.
            (VLMath.h)
        changed headers to have vl/ prefix (unix version only)
        renamed vl_finite to vl_is_finite
        added vl_inf to VLMath.h
        added vl_rand to VLMath.h
        added clamp[ed] for VecN
        added sign()
        added Factor.cc SVD/QR factorization routines
        updated docs to reflect this
        added xform for 2d & 3d, and mixed versions
        changed makefile structure to current system

1.1.3   Added linear solver documentation, fixed some windows compile 
        issues under VC++, minor fixes to solver code.

1.1.2   Added sub, first and last functions for SubVec, and (re?)added
        row function. Added submat functions for SubMat. Added comments
        about the TReal, TVec, etc. types to VL.h

1.1.1   Fixed problem with += operators and gcc under linux.

1.1     Added oprod operator, irix n32 CC support, changed makefiles & afs
        support, eliminated most g++ warnings.