should there be reflexive paths?

schmerl / LLStaging

Staging area for releases to Lincoln Labs

2 stars 0 forks source link

should there be reflexive paths? #61

Open ivoysey opened 7 years ago

bisc commented 7 years ago

More specifically: what should be the response of the base case when told to go from l1 to l1? Is it appropriate to fail?

ivoysey commented 7 years ago

well, really not just l1, right? the question is "for all waypoints w, what should the response be when told to go from w to w?"

bisc commented 7 years ago

true, that was an example. I actually discovered it on l6

schmerl commented 7 years ago

Here is the logic I've intuited that LL is using:

Run the base case without any perturbation. If that succeeds
Run the base case with the perturbation. If that fails,
Run the adaptive case. So, what you're saying will fail the first one, in which case LL will say "Oh, that's not a good test case"

bisc commented 7 years ago

In my view, that's their automated procedure to approximate what is a good test is. One could argue that a corner case is also a good test (but perhaps not a very interesting one).

It's strange for us to manipulate their procedure into excluding tests that are a priori good.