edge case: responsibility for slices > 1/2 key space #48

New issue

Closed

opened 2020-06-10 17:11:11 +02:00 by schmittlauch · 0 comments

schmittlauch commented

2020-06-10 17:11:11 +02:00

Owner

In 1204457a2a , filtering of predecessors and successors has been introduced to enforce predecessors/ successors being smaller/ larger than the node itself.

While this is generally a good idea, it creates problems when the first real predecessor node is more than half the key space size away, making the comparison returning "larger" and the predecessor not being put into the predecessor list.

Example:

Node1 = 1000
Node2 = 1003
Node3 = 1010
Node4 = 1020

with k = 2 Node1 has the successors 1003 and 1010, but no predecessors.

TODO:

Is the responsibility correctly assigned, returning FOUND for the case of empty successor?
Not being able to stabilise with that node, will this sooner or later rip a hole into the namespace?
- Node4 won't have Node1 in its successor list as well
- write tests and decide

In 1204457a2adc32bf28fd00925cea5658466b20a2 , filtering of predecessors and successors has been introduced to enforce predecessors/ successors being smaller/ larger than the node itself. While this is generally a good idea, it creates problems when the first real predecessor node is more than half the key space size away, making the comparison returning "larger" and the predecessor not being put into the predecessor list. Example: Node1 = 1000 Node2 = 1003 Node3 = 1010 Node4 = 1020 with k = 2 Node1 has the successors 1003 and 1010, but no predecessors. **TODO**: - Is the responsibility correctly assigned, returning FOUND for the case of empty successor? - Not being able to stabilise with that node, will this sooner or later rip a hole into the namespace? - Node4 won't have Node1 in its successor list as well - write tests and decide