-
guzzi
guess it is time to get back into Monero. Spent a couple years doing my own thing on the side and learned a lot.
-
guzzi
c++ is callin my name again
-
dEBRUYNE
Welcome back!
-
sethsimmons
vtnerd: not sure if you've seen or not, but BIP324 may be relevant to your work on e2ee of p2p comms:
gist.github.com/dhruv/5b1275751bc98f3b64bcafce7876b489#Abstract
-
sethsimmons
Initial associated PR:
bitcoin/bitcoin #18242
-
sethsimmons
Might be useless, but as they're working on a similar protocol change at the same time figured I'd share to be sure it was known 🙂
-
sethsimmons
For anyone who is willing and able to implement this into Monero, I will give a 5XMR bounty for it's inclusion into the codebase via an accepted PR:
monero-project/research-lab #73
-
sethsimmons
This is the best way to get over a key UX hurdle in Monero, and seems to have no drawbacks other than the implementation necessary.
-
sethsimmons
> <@sethsimmons:monero.social> For anyone who is willing and able to implement this into Monero, I will give a 5XMR bounty for it's inclusion into the codebase via an accepted PR:
monero-project/research-lab #73... (full message at
libera.ems.host/_matrix/media/r0/do…4c5ef5aeade87cc89ae9eb7d60ca7499baf)
-
UkoeHB
Anyone have thoughts on SipHash for computing view tags? It was recommended to me
github.com/veorq/SipHash
-
sethsimmons
Already up to 15XMR 😎
-
sech1
UkoeHB it's better to use keccak (cn_fast_hash in Monero codebase IIRC)
-
sech1
no need to add one more hash just for one thing
-
UkoeHB
sech1: the point is it might be way faster, which allows some speedup on view scanning
-
UkoeHB
I will do some tests today
-
sech1
you're optimizing 1% of the total time, the rest is scalar mult
-
sech1
and other crypto operations
-
UkoeHB
it might be 3%
-
UkoeHB
so, I will test
-
UkoeHB
since this SipHash is very trivial to use
-
sech1
is it cryptographically secure?
-
sech1
ok, it is
-
sech1
we also have Blake (in RandomX), it might be also faster
-
sech1
*Blake2
-
sech1
since you'll be testing performance, it makes sense to test everything we have in the codebase already
-
UkoeHB
I will take a look
-
selsta
.merge+ 8029 8030
-
xmr-pr
Added
-
hyc
so for anyone curious ... I did some tests with LMDB on NFS, which appeared to work, but ...
-
hyc
it looks like mmap'ing stuff permanently ate up RAM, possibly in the nfs daemons
-
hyc
so the NFS server started OOM killing things even though top showed none of them were big memory consumers
-
hyc
and top didn't show *any* big memory consumers. but 3GB went up in smoke till reboot.
-
aberdeenik[m]
How can monero completely replace cash, what property it should be given such that it becomes a full substitute for cash. One area is the fluctuations in price need to be stabilized. Can't it be tethered to say gold reserves?
-
moneromooo
Crash is not tethered to gold reserves anyway. But one of the big problems is reliance on electricity.
-
moneromooo
You can give someone cash if the power's off.
-
moneromooo
Crash is also massively parallel. Your ability to transact in cash does not get lessened if there's a sudden massive increase of cash transactions elsewhere in the world.
-
moneromooo
Cash is also not subject to reorgs and/or 51% attacks.
-
moneromooo
Cash is more private (assuming the serial numbers aren't read as a matter of course)
-
moneromooo
I see you asked the same thing in -lounge, that is a more appropriate venue indeed.
-
aberdeenik[m]
Offline moneros can be generated right , like having a dedicated wallet for each monero and then giving it out as a card
-
aberdeenik[m]
Which has the privatekey
-
aberdeenik[m]
So instead of doing transaction on network we just send the card which has the private key?
-
UkoeHB
sech1: is it `blake256_hash()` that you want me to test?
-
moneromooo
Probably blake2b_{init,update,final}
-
sech1
int blake2b(void *out, size_t outlen, const void *in, size_t inlen,
-
sech1
const void *key, size_t keylen);
-
UkoeHB
oh found it
-
sech1
in monero\external\randomx\src\blake2\blake2.h
-
sech1
the last 2 parameters can be 0
-
UkoeHB
how do I include this into a test?
-
moneromooo
See tests/performance_tests, make a new file (copy from, eg, cn_fast_hash.h), replace with your hash.
-
moneromooo
Then call from main.cp
-
moneromooo
If you want to call the whole wallet2 thing, it's more complex I guess.
-
UkoeHB
I just want blake2b...
-
moneromooo
Then: See tests/performance_tests, make a new file (copy from, eg, cn_fast_hash.h), replace with your hash.
-
moneromooo
subaddress_expand.h does call wallet2, you can start off that if you want to call the "is this output for me" thing with various hashes.
-
UkoeHB
no I know how to do all that
-
UkoeHB
I guess blake2 is not exposed to users of randomx
-
UkoeHB
Ok I have perf tests running. Not sure what to do with them.
-
UkoeHB
-
UkoeHB
On my machine: siphash 3.7us, cnhash 4.9us, blake2b 4.1us, view-scan ~72us. So cnhash -> siphash is about 1.5% faster, and cnhash -> blake2b is about 1% faster.
-
UkoeHB
view-scan with 100% view tag failures*
-
atomfried[m]
is this on your machine? maybe it would be nice to test this speedup on ARM to verify mobile devices and wallets will also profit
-
selsta
.merge+ 8020 8017
-
xmr-pr
Added
-
selsta
8021, 8023, 8016, 7997 need a review, the first three are easy and the last one is just backporting
-
UkoeHB
atomfried[m]: go for it
-
atomfried[m]
i have a raspberry pi with an ARMv8 do you think this will be comparable to mobile phones?
-
moneromooo
tbh, AIUI siphash is well used/tested on bitcoin so appears a safe enough choice.
-
moneromooo
I guess it'd need a cryptographer to check whether it'd be the same properties that get exercised in both uses.
-
UkoeHB
atomfried[m]: idk, maybe? lmao'
-
moneromooo
Most rpi do not have hardwaes AES, and thus suck at PoW verification.
-
moneromooo
I don't know which variants might have HW AES. Maybe yours does.
-
atomfried[m]
hmm ok i guess it would be worthless then
-
atomfried[m]
i will check
-
moneromooo
For mining, certainly useless.
-
vtnerd
sethsimmons : I was (mostly) aware of that BIP - even commented on a variation of it years ago on bitcoin-dev
-
vtnerd
or wait this is a different BIP than what I saw previously
-
vtnerd
a lot of the BIP is working with the existing Bitcoin protocol, which differs from the Monero protocol