pull down to refresh
110 sats \ 5 replies \ @BITC0IN 19h \ parent \ on: Where does the filtering stop? bitcoin
I mean that depends. part of the problem is peoples ignorance surrounding this issue.
disclosure seems prudent to me.
fyi: despite that blocksonly code you reference being in place, it does not functionally work when I tested sending bitcoin in the field with my own node. there may be other factors at work though, i did not test this extensively.
another draw back of being blocksonly=1 is that my peers (i'd say most) start to disconnect from me, since they think I'm a bad peer.
it does not functionally work when I tested sending bitcoin in the field with my own node
I remember it used to be like that - with 0.12 or so - but it should be fixed now. iirc the main issue is that since
smartfees tracks the number of blocks it took for a tx to be mined since we saw it, there is no good fee estimation (though this could be fixed.)I run a blocksonly node to serve history for others' IBD over tor, but I've compiled it w/o wallet so I cannot test it quickly right now (and I don't really like testing on mainnet anyway). I'll build and run some reproducible tests tomorrow, because it should work (as permissioned txs should be forwarded as well, so I'll test that too) - and if not, I think that it's patch-worthy - I'd gladly work on a fix if this was regressed (and with it supply a test with higher coverage.)
my peers (i'd say most) start to disconnect from me
There are 2 blocks-only slots and 8 non-blocks-only outgoing connection slots per node, so you'll serve a smaller subset of total connections, yes, and it'll take longer to serve many incoming peers.
reply
iirc I only had one consistent peer when being blocksonly=1 when trying to sync to the chaintip. once at the chaintip I think i got more peers, but still very few.
Thanks for looking into this, me and the one other user in the world who used this feature appreciate it.
oh also, there's some edge cases with lightning nodes who switch blocksonly=1 on too. terrible cases of channel opens being stuck in some instances (just me lol) that required LND updates to flush out.
also,
put me in the release notes! ;D
reply
reply
If LND now supports blocksonly then that's an awesome improvement that I didn't know worked. I remember needing a mempool for it to work in the past.
no I don't think it does work, which is what I found out the hard way with some stuck channel opens.
Might be worth it for LND to disable that argument actually, now that I think about. it's a foot gun that I used well to clean the toes of my node.
go make a PR and claim credit!