o The path itself isn't stroked, just the fill.
o Lots of elements not implemented fully or at all.
o libsvg's use of a length structure or a double as a passed argument isn't
  clear, or at least I don't understand it.
o Scaling of length variables probably isn't correct.
o I hack into rgb values of the libsvg color or rgb structure, whatever it's
  called. libsvg provides inline functions to get the r, g, b and a values
  I think, these are probably safer.
o Perhaps replace floating point with fixed point integer/fractional values?
  I was originally intending this when I named the point structure "IPoint".
  Then I ended up using floats.

----------------------------------------------------------------------------
o DONE Sun Jan 16 08:27:23 PST 2005
  Fix problem in elements composed of multiple paths, like the letter 'O' which
  would have an inside path and an outside path. The spanning algorithm
  fails miserably on this right now. The issue is that the "sense" of each
  row needs to be consistent for all paths, but it is only consistent for
  a single path. That's because left to right vs right to left crossings
  matter, and there is no guarantee the inside path starts on the same row
  as the outside path (in the 'O' example).
  This might be the cause of the path failures in the svg/bored_dog_01.svg file.
  Or it could be just edge effects on the end or beginning of a path.
  This really would only be a problem on the winding mode for fills.
o DEFUNCT Thu Feb 17 18:01:28 PST 2005
  Span algorithm needs cleanup on start/end of path, it's pretty cryptic and
  I'm not sure if it is completely correct.
o DONE Thu Feb 17 18:01:28 PST 2005
  No attempt at anti-aliasing
o DEFUNCT Thu Feb 17 18:01:28 PST 2005
  Optimization is possible in a lot of rendering code. For example the
  CurveTo function converts all bezier curves into 16 LineTo's. Maybe this
  is overkill. Or maybe it's not enough when the curve is magnified.
  The inner loop of the radial gradient can be tuned a little.
o DONE Thu Feb 17 18:01:28 PST 2005
  "solid" function generates a compiler warning. I left them in since it's
  a work in progress.
o DEFUNCT Thu Feb 17 18:01:28 PST 2005
  Span algorithm has fixed maximum sizes, this isn't really wise.
o DONE Fri Feb 18 17:49:08 PST 2005
  The alpha channel isn't handled as fully as I'd like. For example suppose
  you're rending SVG into an SDL surface, the intent being the SDL_surface
  itself would be applied to another SDL_surface, with an alpha channel.
  The resulting alpha channel on the SDL_svg rendered surface probably won't
  contain the correct values.
o DONE Sun Feb 20 20:25:32 PST 2005
  autoconf would be nice. Enzo's project might have had one.
o DONE Sat Mar 19 17:45:29 PST 2005
  Fill coordinate system isn't correct. For example in radial gradients the
  implementation always draws perfectly circular gradients. The gradients
  in that case need to be oval shaped depending on the bounding box of the
  element (I think?) I think all the gradients need x + y scaling in some
  form.
o DONE Sun Mar 20 19:59:04 PST 2005
  Rectangle implementation doesn't handle rounded corners
