#pest | 2023-07-16

looks like muh trb node took just under 2mos to sync

09:20

recent ryzen, beefy m2 ssds, ddr5, decent box.

09:21

I departed from tradition in a few ways. I did end up snipping out malleus, as I'm suspicious that it bans just about all prb these days, rather than just thin nodes, or whatever they were called, spv

↖ ↖

09:21

trbism does not rule the bitcoin network, so I don't think it's worth it anymore to try and detach from all nodes that issue an unknown command.

09:21

I just went back to silently ignoring them.

09:22

I also set the version to 0.5.3, and that seemed to increase sync rate.

09:23

this node also had my wallet snip patch, but I doubt that improved performance, just cleaned up the code.

09:23

also snipped out anything to do with tx handling during sync.

09:24

*

signpost has often seen a syncing node get lost trying to connect transactions it has no business handling until chain's topped up.

09:24

not doing hard science over here; just some observations

09:26

baking flags such that these (aside wallet, which I have never used nor want) can be flipped on and off probably wouldn't hurt so it could be understood better.

09:31

asciilifeform: do you still want to keep cement out of "mainline" trb

09:31

(asciilifeform) 2022-02-19 asciilifeform: but asciilifeform would rather leave the decision of 'cement or not' to indiv. operator, rather than try to make it hardcoded piece of trb.

09:32

or did I misunderstand? iirc still marked draft on your blog

~ 1 hours 38 minutes ~

11:10

awt_akris

!!ticker btc usd

11:10

deedbot

$30,447

~ 1 hours 11 minutes ~

12:22

http://logs.bitdash.io/pest/2023-07-16#1028350 << maybe this can be narrowed to commands which specifically indicate a parasite pseudonode, vs a full node issuing prbisms

12:22

Logged on 2023-07-16 09:21:29 signpost: I departed from tradition in a few ways. I did end up snipping out malleus, as I'm suspicious that it bans just about all prb these days, rather than just thin nodes, or whatever they were called, spv

~ 3 hours 28 minutes ~

15:51

http://logs.bitdash.io/pest/2023-07-16#1028348 << was this all-natural sync or w/ cement ?

15:51

Logged on 2023-07-16 09:19:42 signpost: looks like muh trb node took just under 2mos to sync

15:51

http://logs.bitdash.io/pest/2023-07-16#1028350 << iirc a # of folx did this (incl. asciilifeform , who not found a significant effect), prolly it oghta be a flag

15:51

Logged on 2023-07-16 09:21:29 signpost: I departed from tradition in a few ways. I did end up snipping out malleus, as I'm suspicious that it bans just about all prb these days, rather than just thin nodes, or whatever they were called, spv

15:53

http://logs.bitdash.io/pest/2023-07-16#1028353 << would be surprising to asciilifeform if this reliably helped ( other than via malleus-like rejection of egregiously useless recent prb peers -- iirc prb after certain version simply drops versions prior to certain # )

15:53

Logged on 2023-07-16 09:22:50 signpost: I also set the version to 0.5.3, and that seemed to increase sync rate.

15:54

ah, maybe I unmalleus'd in one place, then implicitly re-malleus'd with this.

15:54

fwiw per asciilifeform's measurements, virtually all sync time in an all-natural (sans cement) sync is spent waiting pointlessly for sumbody to actually send $next-desired-block

15:54

signpost: likely

15:55

http://logs.bitdash.io/pest/2023-07-16#1028359 << asciilifeform is satisfied that it worx ( see cgra's comments tho , could use certain minor cleanups )

15:55

Logged on 2023-07-16 09:31:48 signpost: asciilifeform: do you still want to keep cement out of "mainline" trb

15:55

was a non-cement sync, but did -connect= to someone's trb node towards the end. got to about 5-600k before doing this iirc

15:55

(asciilifeform) 2022-02-19 asciilifeform: but asciilifeform would rather leave the decision of 'cement or not' to indiv. operator, rather than try to make it hardcoded piece of trb.

15:58

( ftr ~100% of sync time ~with~ cement is spent on verifying blox. which, if one were using cement generated personally, and a trb press known to otherwise work, could safely skip; but asciilifeform was not comfortable encouraging this practice and so not included the requisite toggle in the draft )

15:59

^ if were going there, would make over9000x moar sense to fetch the cement blox in parallel from as many peers as possible and ~then~ verify'em

16:03

http://logs.bitdash.io/pest/2023-07-16#1028351 << erry time asciilifeform stands up a noad, is surprised that sync still worx at all , considering the ecology of current net

16:03

Logged on 2023-07-16 09:21:54 signpost: trbism does not rule the bitcoin network, so I don't think it's worth it anymore to try and detach from all nodes that issue an unknown command.

16:06

*

asciilifeform not burns with desire to reopen the hood on trb (barring discovery of catastrophic bugola), as errytime he does so quickly comes to conclusion that what it really needs is a jerry can of gas, a bic, and a sane rewrite

16:06

(asciilifeform) 2020-10-01 asciilifeform: gregorynyssa: there's not a thing that'd rise to the level of 'plan'. asciilifeform in particular wrote 3 things: 'nqb', a largely-complete coder/decoder for the formats used in trb; ffa, with which possible to

16:06

( sadly asciilifeform not has resources for the latter atm or in foreseeable future )

16:06

yep, I'm just doing things to it that make it usable for my purposes, like carving a sane txn hole in the side.

16:07

(asciilifeform) 2021-10-26 asciilifeform: sadly not had chance to weld 'cryostat' to 'nqb', got into an unspeakably arduous commercial contract soon after publishing the former, and still mired in liquishit

16:07

a++

16:09

*

asciilifeform does find the lack of any detectable, with naked eye, commercial interest in a sane bitcoin implementation, rather peculiar

16:09