WIP feat/routing: use node key as name

[maqi]

This change is

[maidsafe-highfive]

r? @afck

(maidsafe_highfive has picked a reviewer for you, use r? to override)

[afck]

Review status: 0 of 7 files reviewed at latest revision, 3 unresolved discussions, some commit checks failed.

---

src/xor_name.rs, line 126 [r2] (raw file):

        let mut distance = XorName([0u8; XOR_NAME_LEN]);
        for i in 0..XOR_NAME_BITS {
            distance.0[i] = self.0[i] ^ target.0[i];

That's not minus, that's xor. But for the interval length computation we'll need minus. Or rather: the absolute value of the difference. E.g. compute self - target if self >= target, otherwise return target.minus(self) instead.
And maybe it should be called abs_diff then, or euclidean_distance.

---

src/states/node.rs, line 979 [r2] (raw file):

        println!("input nodes {:?}", nodes);
        let name = XorName([0u8; XOR_NAME_LEN]);
        nodes.sort_by(|node0, node1| name.cmp_distance(&node0.name(), &node1.name()));

Doesn't XorName implement Ord anyway? (If not: it should, I think.) So you could probably just call nodes.sort().

---

src/states/node.rs, line 983 [r2] (raw file):

        let mut max_range = XorName([0u8; XOR_NAME_LEN]);
        let mut range_index = 0;
        for i in 0..(nodes.len() - 1) {

Once Disjoint Groups are in place, this will need to include the lower and upper bound of the group's range (if there are no nodes with exactly those names).
Let's either add a TODO or hold off with merging this PR until after RFC 37 is implemented.

---

Comments from Reviewable

[afck]

Review status: 0 of 7 files reviewed at latest revision, 3 unresolved discussions, some commit checks failed.

---

src/xor_name.rs, line 126 [r2] (raw file):

Previously, afck (Andreas Fackler) wrote…

That's not minus, that's xor. But for the interval length computation we'll need minus. Or rather: the absolute value of the difference. E.g. compute self - target if self >= target, otherwise return target.minus(self) instead.
And maybe it should be called abs_diff then, or euclidean_distance.

Actually, ignore the "absolute" part of my comment: Just implementing subtraction should be fine, since you sort the names below anyway, so you always subtract the smaller from the larger one. I think we should just `impl std::ops::Sub for XorName` then: implement subtraction for 256-bit integers.

---

Comments from Reviewable

[maqi]

src/states/node.rs, line 979 [r2] (raw file):

Previously, afck (Andreas Fackler) wrote…

Doesn't XorName implement Ord anyway? (If not: it should, I think.) So you could probably just call nodes.sort().

we can call nodes.sort(), but it doesn't change the order (or it sorted in the decreasing order?)

---

Comments from Reviewable

[afck]

Review status: 0 of 7 files reviewed at latest revision, 3 unresolved discussions, some commit checks failed.

---

src/states/node.rs, line 979 [r2] (raw file):

Previously, maqi wrote…

we can call nodes.sort(), but it doesn't change the order (or it sorted in the decreasing order?)

Ah, sorry, it is `PublicId` that needs to implement `Ord` - and that should definitely go by _name_. Maybe it will, if we just make `name` the first field in `PublicId`? If not, `Ord` needs to be implemented by hand.

In either case, please add a test to id.rs to make sure that public IDs are ordered by name.

---

Comments from Reviewable

[dhardy]

I don't follow... why would Z have A's original ID? For the previous line, it would be nicer if your addition was spelled out: "node Z (of NaeManager Y)"

[maqi]

the group X has notified group Y about A's original id

[dhardy]

Spelling: "Disjoint", "guaranteed"

[dhardy]

Lets hope the client/joining node is not busy, or it might not manage many iterations. Often this type of code allows a maximum number of tries instead of a timeout.

[maqi]

the network side is waiting based on timeout, so here needs to be wait on timeout as well.

[Viv-Rajkumar]

closing this for now. lets get this raised and merged when switching the node name pattern with node ageing.