decred / dcrseeder Goto Github PK

The seeder does not send EDNS flags correctly in the responses, this may cause issues when using some recursive name servers: https://dnsflagday.net/#domain-holders

EDNS compatibility report:
https://ednscomp.isc.org/ednscomp/d6e17b0470

IP addresses that are not publically routable should be ignored and never added to the cache.

The value is currently hardcoded.

New protocols generally only add features, so it is more useful to query for the minimum version required instead of the exact version. This allows older software to still reach updated nodes with seeded addresses.

perhaps a configuration option to display statistics every specified unit of time.

number of live nodes grouped by services
total number of live nodes:
number of IPs that still need checking:
number of dns lookups

etc...

Need to remove addresses from the cache that have never worked or haven't worked in some specified duration.

when exec dcrseeder -n seedserver.xxx.xxx -H mainnet-seed.xxx.xxx -l 2001:ac8:40:xxxx::xxxx:xxxx:53 -s 2001:ac8:40:xxxx::xxxx:xxxx
got a problem ResolveUDPAddr: address 2001:ac8:40:xxxx::xxxx:xxxx:53: no suitable address found
So now only support IPv4?

The data folder used to store nodes.json is always ~/.dcrseeder. This should have the network name in it too (e.g. ~/.dcrseeder/testnet3).

documentation is needed on how to setup the DNS zone file.

Ideally a single dcrseeder process should be able to crawl both mainnet and testnet:

• Specify one or two hostnames (one or both of mainnet and testnet)
• Crawl one or both networks simultaneously
• Launch a single DNS server that resolves one or more host names (two in the case of mainnet and testnet)

e.g.

• Name server: ns.blah.org
• Testnet host name: testnet-seed.blah.org (with NS record pointing to ns.blah.org)
• Mainnet host name: mainnet-seed.blah.org (with NS record also pointing to ns.blah.org)

currently "about" says " Decred DNS Seeder " change to "Decred Seeder"

It has been a long time since I've looked at the old seeder code but I know that it did not require using a running node for it to run. It would be best if this seeder could do the same.

I do remember that the old seeder had a hardcoded list of all the other seeders:
https://github.com/decred/decred-seeder/blob/19351f3e3ec93b96ca3438c465c04163747d0c5c/main.cpp#L400
Did it use that to get the initial node list (assuming at least one of those must be up and possibly serving a static list of 'good' nodes, at least during the initial network bringup?

Regardless of the exact method, require a full node is a lot to ask for running a seeder (which is otherwise a very light task).