< sipa>
always surprising to see these large powers of 10 as factor appear
< sipa>
around 10% of those are len512 hd5
< sipa>
so the next question is how to prioritize hd5 length vs hd6 length
< sipa>
without ability to reach hd6 len222 at all, we should probably just maximize hd5 length
< gmaxwell>
Agreed.
< gmaxwell>
It might be interesting to see what we end up with under each criteria? what do we get prioritizing hd6 length, then hd5 lenght, what do we get prioriting hd6 length then hd5 lenght?
< bitcoin-git>
[bitcoin] amitiuttarwar opened pull request #15332: [Docs] Small updates to getrawtransaction description (master...get_transaction_docs) https://github.com/bitcoin/bitcoin/pull/15332
< sipa>
gmaxwell: longest hd6 is len131
< sipa>
not enough for 2 public keys even
< sipa>
that one is len788 hd5
< phantomcircuit>
wumpus, enums for error codes is definitely the way to go
< sipa>
gmaxwell: longest hd5 is len1318
< sipa>
which is only len22 hd6
< bitcoin-git>
[bitcoin] MarcoFalke closed pull request #15206: Immediately disconnect on invalid net message checksum (master...019/01/netmsg_2) https://github.com/bitcoin/bitcoin/pull/15206
< bitcoin-git>
[bitcoin] MarcoFalke reopened pull request #15206: Immediately disconnect on invalid net message checksum (master...019/01/netmsg_2) https://github.com/bitcoin/bitcoin/pull/15206
< sipa>
gmaxwell: what do you think about letting the existing descriptor-capable RPCs (scantxoutset and deriveaddress) accept both descriptors with and without checksum (but if a checksum is present, it must be correct), but enforcing checksum for future importing RPCs
< gmaxwell>
sipa: scantxoutset is probably okay, but deriveaddress sounds dangerous.
< gwillen>
agreed about deriveaddress I think? Certainly it ought to warn. Deriving an address is the most dangerous operation.
< sipa>
right, agree
< sipa>
there is no recourse if you get that step wrong
< sipa>
so my thinking is i can probably analyse all 5797 deg8_len512hd5 codes up to length 300 for up to 6 errors, and then pick the one that performs the best at that length (300) for say 1% errors
< bitcoin-git>
[bitcoin] hebasto opened pull request #15334: wallet: Log subdirectory for the named wallets (master...0190203-wallet-env-log) https://github.com/bitcoin/bitcoin/pull/15334
< sipa>
300 is enough for a k-of-3 wsh/multi with hex pubkeys, where the class symbols are inline
< promag>
MarcoFalke: regarding your comment, could you open an issue with more details of your idea?