Clarify that the path selection should be better than not selecting a path

[renghardt]

Fix #87 - The default case is that we don't do path selection at all. What exactly happens then will depend on the context, but I'd say this usually means that a single "default" path is used, which may be statically configured. But I don't think it's necessary to spell this out here. I am pretty sure that we don't want to compare it to a random selection, because I've often seen cases where the default path is "good" and the alternative path is just unsuitable. In such a case, if you select the path randomly, you'll perform much worse than the default of not doing path selection at all.

[renghardt]

As with the other PR, I added Med and Dave in case they want to chime in as well. If I don't hear anything within a week (by March 3rd, 9am PT) I will merge.

[daveoran]

I'm ok with this, but I'll try one last time to suggest that "perform better" doesn't capture the essence of why a particular path might be selected by a user. As I tried (i guess unsuccessfully) to get across, a user might intentionally select a path that performs worse (in terms of the usual "performance" measures like bandwidth, latency, etc.) in order to satisfy some other objective function they have, like lowering cost, or increasing the probability that some other path selection request will succeed for other traffic they generate. That's why i suggested the wording "…a path selection algorithm should aim to do a better job of accommodating the user’s requirements than the default case of not selecting a path”.

[renghardt]

Ohh, I got you now, thank you for reiterating this aspect of your comment. I agree that the objective is not always related to performance, and I changed the text to reflect that.

Everyone, please let me know if the proposed phrasing sounds good to you.