Where can i find the logo of the research lab?

nevermind. I downloaded the logo of this room on matrix

ErCiccione[m]: rehrar might have some high-res versions too

Also, hello all

Finalizing common-key Triptych batching and getting the plots updated today, woo

🎉

that sounds awesome!

On a desktop/laptop does validating two blocks in parallel take the same amount of time as validating two blocks serially?

Probably not.

The concept of "validating two blocks in parallel" is ill defined though.

I'm booting up a new node. Instead of processing blocks 1 - 1000 in a row, perhaps I validate [1-250] in parallel with [...], [...], and [751-1000]

those example numbers aren't so great, since monerod uses compiled-in hashes for the first couple million blocks

but since blocks refer to previous txns, you can't get very far ahead, if at all, by parallelizing

Blocks are made of txes, which depend on previous txes, so it might be annoying to parallelize, but by all means, any improvement is probably good :)

if you're lucky enough that all the txs are old enough to have already been verified

if not, then you're implicitly serialized anyway, waiting for those txs to go by

Oh very interesting, I have to think about that

You can verify PoW in parallel.

Without dependency on earlier blocks I mean.

does predictive/optimistic branching/rollback make sense in this context? i.e. groups of blocks expected to be valid done in parallel, invalid triggers rollback + serial validation

it's a nice idea in a perfect world

because in the ideal case, it would work in parallel most of the time

but if you become dependent on that, it's trivial to DoS the network by tossing in crafted blocks

true, could do nasty things w/ invalid blocks nested deep in a valid chain... damn

what about using some of the snapshotting tech to guard against maliciousness? so, snapshot 100k block chunks, do a hash/zkp of the snapshot, validate all included snapshot tx in parallel

or w/e for the snapshot # of blocks

A zkp of what exactly?

(I only ask because my eye twitches whenever I hear "just use a zkp" for something)

doesn't verifying PoW depend on recomputing all the difficulties, which is a serial operation?

zkp doesn't make sense

disregard, but the hash of the snapshot as checksum

derp dropped (irc gods are angry w me)

derbleak: monerologs.net/monero-research-lab/20200421 if you need to recover logs

UkoeHB_: thank you :)

re: binary distribution, one possibly dumb idea could be to use serverless functions to `git pull and build`, distributing a freshly built bin every time :)

be fun to see amazon reaction

:p

bezos trolling aside, think the torrenting idea is great

bittorrent has a blockchain project on tron iirc, could enable another channel of decentralized hosting/distribution

pls no tron

lol 100% ^

mb filecoin or smth similar?

did filecoin ever launch?

long way to say no

yeah...

sorry, posted b4 reading

ipfs is definitely live though

siacoin is another one, think storj went out of business(?)

or whatever happens when blockchains die

bits used as entropy, fading into the heat death

a bit off topic here but what would the advantage of torrent / ipfs / ... be over a normal http download?

security and baked in authentication/signing

for ipfs, potentially censorship-resistant, *very* long-term storage

similar for siacoin etc.

could also host the storage coin clients as hidden services, mb

s/clients/nodes

mb i'll set one up today :)

New multi-input timing data: usercontent.irccloud-cdn.com/file/8CJqeZGu/timing.png

This example is the per-input timing data for a 2-input transaction

Also rewrote some tests for better comparisons that include commitment offsets

which version of triptych is that?

The C++ implementation I did

with some modifications to support commitment offsets more efficiently

Shortly, I'll do a quick modification to support multi-index signing and balance assertion

triptych or arcturus

Triptych

^ is this data

Arcturus should have similar timing results

thx

with better size data