Updated Difficulties in use in README.md

[shingtaklam1324]

Difficulty three is in use by saddle-points and isogram

[coriolinus]

I believe there also exist come difficulty 6 exercises.

Is it worth continuing to enumerate all difficulties in use? If we have a ten point scale, in my mind, we may as well use all ten points.

[petertseng]

Yeah too bad, the cat's been out of the bag for a long time. That's really too bad because 1 4 7 10 was the easiest way to transition out of https://github.com/exercism/rust/blob/aea470a76a9c6f0005adebb2a00d969512da3b6e/problems.md and I hate having to AP around which difficulty a problem should be. Shrug.

[coriolinus]

$ jq '.exercises[] | .difficulty' config.json | sort -n | uniq -c
      2 null
     18 1
      2 3
     45 4
      2 6
      4 7
      8 10

We currently use 6 of the 10 possible difficulties, the vast majority of which are either 1 or 4. This suggests two things:

1. We should update the readme in one of three ways:
   • Enumerate all current difficulties: 1, 3, 4, 6, 7, 10
   • Adjust the level 3 exercises such that they are level 4, and adjust the level 6 exercises such that they are level 7, restoring the 1 4 7 10 separation.
   • Rephrase to eliminate the enumeration, and have people just assign a plausible difficulty to their exercises.
2. We should give some thought to increasing the number of challenging exercises in this track. They're the fun ones, once someone slogs through all 60+ exercises at and below difficulty 4.

Of the possibilities in point 1, I prefer the last, though I could be talked into supporting the middle if there's a strong argument in its favor. The first seems suboptimal to me.

[petertseng]

I can't really justify listing 1, 3, 4, 6, 7, 10 in the README since there doesn't seem to be much rhyme or reason in why those numbers are included and 2, 5, 8, and 9 are excluded. 1, 4, 7, 10 made sense because it advanced from 1 to 10 in equally-spaced increments, and it was easier to make a decision when there were fewer choices. But I should have made Travis CI enforce that if I wanted that to remain the case.

Using more difficulties may help us understand how to order exercises better. I can't guarantee I'll spend a lot of time thinking where the boundaries between any two adjacent numbers should be, but I'll probably notice if an exercise is rated much too low or high.