< gmaxwell> Github's goal, to make people hate unicorns.
< luke-jr> lol
< dongcarl> Have there been any thoughts put into an identicon/visual hash representation of Bitcoin addresses for improved user experience? Would that be useful in any way?
< gmaxwell> dongcarl: prior work on visual hashes have suggested that they're easily forgable. Users also generally should not be reusing addresses, so 'remembering what the right one looks like' doesn't really apply.
< gmaxwell> (e.g. there is a paper that forges look-alike ssh visual hashes)
< gmaxwell> a long time back I did post something about bosting fingerprint comparison security via cut-and-choose: https://en.bitcoin.it/wiki/User:Gmaxwell/visual_fingerprint_comparison
< dongcarl> gmaxwell: Wouldn't there be other cases where one would want to easily verify that the typed in address is the one that is intended other than 'remembering what the right one looks like'?
< * dongcarl> reading
< luke-jr> why not just use a QR Code? <.<
< dongcarl> I mean something like this: http://identicon.net/
< wumpus> phantomcircuit: there were terrible races around closing sockets in the past, wouldn't surprise me if satoshi had tried to hack around it that way
< dongcarl> Like what Github uses for default gravatars
< dongcarl> the color adds a dimension too
< gmaxwell> We know what identicons are.
< luke-jr> point is if you have a visual, you can just skip typing it in in the first place
< gmaxwell> dongcarl: you're going to have to be more specific about "Wouldn't there be other cases" ... like what, specifically? if you're worried about typos, the addresses have phenomially strong checksums that preclude typos.
< gmaxwell> phenomenally*
< dongcarl> You're right
< dongcarl> gmaxwell: the hexagonal cut-and-choose fingerprint you proposed isn't needed either because of checksums?
< jonasschnelli> is there an easy way to serialise a 3 byte int? With things like CVectorWriter?
< jonasschnelli> Splitting into 3 unit8 seems to break the idea that the serialization layer takes care of LE/BE
< luke-jr> we kinda do it for height..
< jonasschnelli> luke-jr: I can't find it in the code,... can you point me to it?
< luke-jr> it's the script serialization stuff
< jonasschnelli> okay... I'll look into it. Thanks
< echeveria> dongcarl: SSH has had visual host keys for years, but I'm not sure anybody has ever seriously used them.
< wumpus> I've closed some older PRs that seemed to be inactive for a long time, so as always: let me know if you start working on something again and I'll reopen
< wumpus> "visual representation of addresses" is one of those ideas that comes up often, but no one ever has been motivated enough to write a BIP about
< echeveria> I'm not really sure there's a great way to do it
< echeveria> blockchair.com does a particularly terrible job of it.
< wumpus> I'm not really convinced it's very useful either, but apparently there's a subset of people that does, seeing people talk about it every so often
< wumpus> agree with luke-jr that QR codes are the way to go
< wumpus> at least they can be usefully scanned, and you can generate them everywhere, even on the terminal with unicode block characters
< ken2812221> if I pass "bitcoind -nodebuglogfile=0", it will log to file "1". This is funny.
< phantomcircuit> the rpc_help.py functional test looks for wallet and zmq both of which are compile time options
< phantomcircuit> (indeed zmq is actually default off) so confusingly it always fails for me
< phantomcircuit> that appears to be wrong, the notify is always called with the last size of vNodes not the current size
< phantomcircuit> cant explain why that still seems to work though
< phantomcircuit> sooo
< phantomcircuit> huh no it's right but only on accident basically