WHAT ASPECT OF THE PROBLEM WAS MOST DIFFICULT? 
Secondary color

WHAT INSIGHTS (IF ANY) DID YOU GAIN FROM WORKING ON THE PROBLEM?
1. Don't be afraid to try wacko ideas such as thin shells of glass.
   Because of the all-cemented requirement, I didn't expect it to
   look like anything familiar.
2. One simple idea which I used was to take a familiar lens form
   (I started with a Sonnar, 5 elements or so) and replace the
   air with a low-index glass (FK54), while allowing the original
   elements to have very high index.
3. Don't give up on an optimization just because it converges slowly.
4. Because of the merit function definition, this problem did not
   require anywhere near as much twiddling of field weights as did
   the 1994 problem.

ADDITIONAL COMMENTS:
5. The usual rules about element manufacturability did not apply.
   After all, nobody's going to build it.  But I did keep my
   minimum CT and ET at 1 mm.  Because the lens tended to get
   as long I'd let it (to reduce the angular FOV), this was not
   much of a restriction.
6. I controlled weight by constraining overall length.  Never
   bothered with a complex macro, although I did use macros
   for evaluation purposes.
7. I controlled relative illumination by adjusting vignetting
   occasionally and constraining distortion to be negative.
   At the end, I set vignetting by introducing a few clipping
   apertures, so as to minimize weight while holding the relative
   illumination.
8. I used the GDP feature in Code V to select some KzFS and FK
   glasses.