< GitHub122>
bitcoin/master 37a7fe9 isle2983: [copyright] add MIT License copyright header to zmq_sub.py
< GitHub122>
bitcoin/master 1cf11cc Wladimir J. van der Laan: Merge #8701: [copyright] add MIT License copyright header to zmq_sub.py...
< GitHub99>
[bitcoin] laanwj closed pull request #8701: [copyright] add MIT License copyright header to zmq_sub.py (master...zmq_sub-copyright) https://github.com/bitcoin/bitcoin/pull/8701
< GitHub113>
[bitcoin] MarcoFalke opened pull request #8760: [init] Get rid of some ENABLE_WALLET (master...Mf1609-walletInitGuard) https://github.com/bitcoin/bitcoin/pull/8760
< instagibbs>
sipa: does the v2 cb assume that clients will prefer newer versions over old? I can't make heads or tails of how version agreement is achieved otherwise
< instagibbs>
I'm just reading the bip for now
< gmaxwell>
wumpus: perhaps someone with a running fedora system could confirm, but I believe the maximum locked memory ulimit by default on fedora and centos is 32kb.
< gmaxwell>
wumpus: any idea why we're using so much locked memory? 300k for a big wallet sounds like a lot.
< BlueMatt>
the help documentation for getblocktemplate does not anywhere explain how to get it to spit out a response post-segwit :(
< BlueMatt>
(nor does the segwit-gbt bip, afaict)
< BlueMatt>
luke-jr: ^
< gmaxwell>
BlueMatt: yes it does...
< gmaxwell>
BlueMatt: you want the relevant BIP9 rules, it's described in BIP 145 and BIP 9.
< BlueMatt>
ahh, its mentioned in a one-line below a chart...thats confusing
< BlueMatt>
anyway, its not mentioned in the rpc help text
< BlueMatt>
and I dont really know enough about gbt to trust myself to not confuse the spec adding the "rules" thinggy to the help text
< luke-jr>
?
< BlueMatt>
luke-jr: the help text for getblocktemplate does not explain how to add the "rules" entry so that it will return anything except an error
< luke-jr>
oh
< luke-jr>
it has a link to the BIPs ☺
< BlueMatt>
.....
< luke-jr>
for better or worse, it's not really feasible to implement GBT from less than the BIPs
< * BlueMatt>
goes to write up a 10-minute replacement that only returns what you need to build a block
< luke-jr>
getwork was just removed
< gmaxwell>
BlueMatt: go write a integrated stratum server instead, that hides all this stuff.
< luke-jr>
gmaxwell: libblkmaker already hides it ;)
< BlueMatt>
gmaxwell: true, but every time I ask anyone they also want to kill stratum
< BlueMatt>
though, really, I think if you were to just return essentially the stratum data, but as binary instead of hex, people would be happy
< gmaxwell>
BlueMatt: well it still leaves the protocol utterly unauthenticated which is bad.
< gmaxwell>
but it would certantly be a usability step up.
< BlueMatt>
true, easy enough to add a sig field that just gets appended to the end
< luke-jr>
+1
< * BlueMatt>
goes to find p2p-auth code to figure out how it stores keys
< TD-Linux>
gmaxwell, 64kB on fedora 24
< instagibbs>
BlueMatt: does the v2 cb assume that clients will prefer newer versions over old? I can't make sense of the latest bip text on agreement, although this might be my head cold....
< BlueMatt>
instagibbs: does the sample negotiation not help? the text is dense, but the sample and actual code are pretty readable, imo
< instagibbs>
it seems to assume it, yes, but it's an "example" after all :P
< instagibbs>
in other words, is it assumed I could send "1, 2" and you could send "2, 1" in the initial messages? If so, I just need to sit down and read it again I think.
< BlueMatt>
yes, you should be able to send any preference you want...if its not clear from the spec tell me where i should change it
< instagibbs>
Ok so then that part is clear, but I can't connect dots. I'll try again tomorrow after sleep.