#pest | 2023-05-14

00:34

I think I'll go with the dark theme as well

~ 10 hours 42 minutes ~

11:17

*

signpost doing a trb sync with incoming tx messages (and AskFor tx upon inv containing unknown tx) skipped during initial sync.

11:17

signpost

seeing how fast this can be made.

11:17

signpost

thing seems to spend a silly amount of time gargling txn it can't do anything about until chain's synced.

11:18

signpost

also sadly snipped the heathen command ban out because by now the network's saturated with nodes that issue them. silently ignoring instead.

~ 1 hours 2 minutes ~

12:20

asciilifeform

signpost: have you tried cement ?

12:22

*

asciilifeform found that it worx a++

12:23

asciilifeform

a cement baked & signed by asciilifeform for blox through 723k can be found in the cement maker article.

12:23

*

asciilifeform oughta bake an updated cement. and if anyone has done so already, plox to post

12:25

asciilifeform

signpost: fwiw per asciilifeform's measurements during the 2017 tx spamola deluge, the latter has miniscule effect on sync time (during most of which the thing simply waits for someone to give the requested block. and when the latter arrives, tx aint being eaten anyway)

12:25

asciilifeform

i.e. asciilifeform found tx liquishit to be 'red herring' in re 'wai sync slow'.

12:26

asciilifeform

not that it aint worth to experiment, but 'how long will noad with xyz take to sync' gives largely random number. (what you actually want to measure is 'how much time it sits with mouth open waiting for nth block'

12:27

asciilifeform

( for a 'cemented' sync, that's ~0. at the obvious cost )

12:29

asciilifeform

asciilifeform's test ftr.

12:29

dulapbot

(asciilifeform) 2022-03-28 asciilifeform: 17 feb -- 28 mar.

12:31

asciilifeform

note that a node that is syncing 'in cement' won't propagate tx.

12:32

asciilifeform

( the blox, incidentally, get verified precisely as if 'synced naturally'. the only change is that while 'in cement', trb knows the hash of the next expected block )

12:33

asciilifeform

^ if this weren't so, could sync in a day or two. but then would be just the same as simply copying the db from an existing noad.

~ 23 minutes ~

12:56

*

signpost was observing the thing going back and forth trying to connect transactions for many minutes, prompted getting out the cauterizing laser

12:56

signpost

but yeah, actually forgot about the cement patch too.

12:56

signpost

re: heathen commands, I should probably put a log message in there and see what they are.

12:57

signpost

I hadn't synced a trb from scratch in a long time, so I'll let this one run regardless, but I mean to dust off my walletsnip patch next.

~ 6 hours 20 minutes ~

19:17

awt

It seems like the station will need to getdata all messages before notifying the client that it's safe to fetch the chain. I can't think of a way to present any sort of progress bar because a station can't know when it has reached the root message, other than simply timing out.

19:20

awt

or more accurately, a station cannot calculate progress towards the root. Has to assume root after timeout.

~ 28 minutes ~

19:48

asciilifeform

awt: what means 'safe to fetch the chain' ? ( why wouldja have to wait 'for root' to use the station ? )

19:49

asciilifeform

you plug it in; a message shows up; and the log slowly grows backwards as the thing fetches netchain. during which you can use the station normally, neh

19:49

asciilifeform

(i.e. you can send out a validly chained new msg)

19:50

asciilifeform

the gui certainly should not need the full backlog loaded into it (ugh) to letcha scroll through it (ougha fetch old msgs 'lazily')

19:51

asciilifeform

the only thing user oughta see as the old msgs getdata is, perhaps, the scroll bar 'growing'

19:51

asciilifeform

(tho even this is optional, heathen chats -- e.g. 'slack' -- do without this)