feat: use deferred initialization of the asnStore #3
Conversation
| // If no mapping exists for the given IP, this function will | ||
| // return an empty ASN and a nil error. |
There was a problem hiding this comment.
is there a default cidr prefix for unknown mappings?
There was a problem hiding this comment.
Can you elaborate a little more? I'm not understanding the question
There was a problem hiding this comment.
if 2 ips in the same /64 are looked up in a space that isn't in the mapping, will they get the same empty ASN, or different ones?
when we make a new ASN in response to a query, we should maybe save it in the mapping with some default prefix so that other IPs close to it cluster with it.
There was a problem hiding this comment.
will they get the same empty ASN, or different ones?
We return "" and when this is used in kbucket we just reject ASNs named "".
Note, this is why we're going to have to change your test in go-ipfs to use an ipv6 associated with a real ASN https://github.com/ipfs/go-ipfs/blob/fe9f49aeb71c753b92ed41922c493ff3bf08ef56/test/integration/wan_lan_dht_test.go#L54
aschmahmann
force-pushed
the
feat/defer-init
branch
from
02764ea
to
d8ddb2e
Compare
This is meant to deal with the issue where initialization of the data structure by loading in 50k entries adds a bunch of time to startup. While we should be using better data structures here + serializing/deserializing the structured (instead of unstructured) ASN map at runtime this should at least allow this library to be used async without slowing down the main thread.