tomaszkam

The difference from the previous example is that now instead of transforming `Itinerary` into `vector`, I notionally transform it to `std::vector` and then flatten it. The code to construct view...

I will disagree with the view on the `Relation` and `IndirectRelation` - the relation in the mathematics means any pairing of the values of two domains (note that function is...

Finally, given the other linked issue, I think that the `IndirectlyComparable` and `IndirectRelation` concepts are currently overconstraining the algorithm - to realize the described effects of algorithms we need only...

> TLDR: the prior requirements established that r is type-independent: bool(r(x1, y1)) == bool(r(x2, y2)) must hold when x1 and x2 represent the same platonic value and y1 and y2...

> And yes, I agree. I'm not suggesting that we need to return the common type / reference semantics, I'm pointing out how they linked the required expressions of Relation...

> I'm not suggesting that we need to return the common type / reference semantics, I'm pointing out how they linked the required expressions of Relation together into something coherent....

[P1716: `ranges` compare algorithm are over-constrained ](https://wg21.link/p1716) aims to address the problem.

@ericniebler Yes, the P1716 did make it through LWG in Colonge. I plan to submit an LWG issue for the problem addressed and will link it here.

Actually both, was accepted by LEWG, LWG has not managed to review it.

I think we should differentiate the algorithms, that are guaranteed to always compare values in two ranges in specific order (`equal`, `mismatch`) and the one that is using fact that...