<rbrunner7> Meeting time. Hello! https://github.com/monero-project/meta/issues/1489
<sneedlewoods> hi
<selsta> hi
<vtnerd> Hi
<jpk68> Hello
<syntheticbird> hola
<jberman> waves
<rbrunner7> Ok, what is there to report from last week?
<jpk68> Me: working on reviewing fuzz testing PRs, finally got SAM support merged + submitted follow-up hardening for it, more AI reviews, GUI stuff, etc.
<sneedlewoods> again rebased and fixed conflicts for 9464, 10232, 10233 and 10819
<sneedlewoods> and lots of changes from AI reviews (including one, still WIP, deep review of 10819)
<sneedlewoods> haven't pushed yet, but I can push if anyone wants to have a look, else I'd prefer to finish the reviews, then do some manual testing for things not covered by functional-/unit-/libwallet_api-tests
<jeffro256> Howdy
<rbrunner7> @sneedlewoods: Well, if you have something like a "release candidate" that won't change again at short notice :) ...
<vtnerd> me: lwsf fcmp++ branch has been rebased and updated to work with newest fcmp++ code. have not pushed out the update (still needs testing). might push it out before running tests, that way the community has lateast update regardles
<jeffro256> Implementing feedback from LLM reviews and working on writing postive and negative core_tests for #10916
<jberman> FCMP++ integration PR followups
<rbrunner7> For the record, and for those who may have not seen it yet: Looks like we have a proof-of-concept payment channel implementation based on the brand new relative locks: https://old.reddit.com/r/Monero/comments/1xx0ub2/floppymd_developed_an_experimental_l2_payment/
<sneedlewoods> +1
<AilliA> +1
<jpk68> +1
<rbrunner7> Maybe it's already a net win to get a talented (as I see it) whitehat interested in Monero, regardless of the merit of that software
<thomasbuilds> @jeffro256:monero.social I think this is safe to merge and existing payment processing software won't be left behind (pun intended)
<thomasbuilds> https://github.com/monero-project/monero/pull/11655
<jeffro256> +1
<rbrunner7> How is Monero currently coping with the bigger flow of PRs that we have now? I guess most are quite small and thus not that hard to handle?
<jpk68> A decent amount of the ones from last week (all from one person) were useless and could be closed
<jpk68> In fact, they were closed
<rbrunner7> Ah, ok, didn't notice
<sneedlewoods> the last two weeks I neglected to look into new PRs if I was not requested to look into, to focus on my own PRs
<rbrunner7> Ah, yes, "Uhudsavasindankacanokcu2"
<jpk68> +1
<thomasbuilds> @rbrunner7: Btw there are old PRs/issues that could be closed? 2018-2020 stuff
<thomasbuilds> Haven't checked just mentioning in case for a cleanup
<selsta> thomasbuilds: I'm closing them from time to time
<thomasbuilds> Ok
<rbrunner7> I guess as long as selsta doesn't start to shout things did not yet go out of hand :)
<selsta> my main issue is I don't know what to do with PRs that harden against malicious nodes, they often make the code more complex and don't fix all attack vectors
<rbrunner7> If used as remote nodes for wallet apps, or also quite in general?
<selsta> if used by a wallet yes
<rbrunner7> Hmm. I see how that could lead to difficult trade-offs.
<jpk68> I imagine it might be productive to establish an agreed-upon 'acceptable' attack surface for this, and have it documented
<rbrunner7> On the other hand, the issues may be all over the place.
<jpk68> Could you elaborate?
<rbrunner7> Probably easy to agree that trying to fix things with rings does not merit high priority this late in the "game"?
<rbrunner7> @jpk68: Difficult to find rules for something that is very varied? And where it's maybe difficult to "think ahead" and find categories?
<jpk68> Maybe, I guess so. Though it could also be documented to help those who submit PRs for this, to reduce the number of redundant ones
<rbrunner7> Anyway, I think thanks to AI things develop so fast now that almost every week looks a bit different than the previous one
<sneedlewoods> +1
<jpk68> +1
<rucknium> IMHO there is a safety problem now with all the p2p spy nodes. The usual advice used to be "run your own node", but there are so many spies now. Some new advice is "use a remote node over a Tor connection", but that's subject to malicious remote node behavior, not to mention centralization issues.
<rucknium> I don't have a solution. Just something I noticed.
<rbrunner7> So maybe those spies won't do us the favor to roll out changes that would limit them in doing their "job", i.e. to spy ...
<jpk68> This might be controversial, but why not just 'trust' a reputable remote node (through Tor)? Surely the one run by, say, Cake Wallet or plowsof is less likely to be malicious compared to some random person
<ofrnxmr> @rucknium: this is documented on docs fwiw
<jpk68> Clearly this is not ideal, but still
<ofrnxmr> Rinning your own node w/o proxy is misinformation wrt it being the most private
<ofrnxmr> Its the most decentralized, but far from most private unless you are using txproxy, anoninbound, and proxy
<rbrunner7> At least your own node won't cheat on you and suddenly ask for a fee of 1 XMR?
<thomasbuilds> +1
<plowsof> my nodes are not on hardware i control, same for people renting vps' - who knows what theyre running 
<rucknium> @jpk68: Nodes get overloaded easily if you point everyone at a few of them. SethForPrivacy's node is usually overloaded because it's recomended so frequently.
<ofrnxmr> https://docs.getmonero.org/interacting/remote-nodes/#comparing-your-options
<rucknium> And a trusted node becomes a targeted node IMHO.
<rbrunner7> Not sure I understand the role of proxies here, why would they be so important in achieving good privacy with your own node?
<ofrnxmr> Trusted doesny nexessarily mean public
<jpk68> +1
<jpk68> @rucknium: This is besides the point (on my part), but I think improving wallet RPC sync scaling is very important for this
<ofrnxmr> @rbrunner7: Dandelion doesnt protect any home user who doesnt setup port forwarding
<jpk68> See: Cuprate's blog post, and my read-write lock PR
<rucknium> I agree also with plowsof about not physically controlling the hardware. My public remote node is the same. On hardware that I don't have physical control of or access to.
<jpk68> monerod could potentially be like 20x more efficient in this regard
<ofrnxmr> controversian scandata coming soomtm
<jpk68> https://github.com/monero-project/monero/pull/11143
<plowsof> Rucknium - it was you who stressed this point years ago and i agreed with you 
<ofrnxmr> @ofrnxmr: (my 50 part series that lowers bandwidth and scanning by 9x)
<rucknium> plowsof: By agreeing with you, I am agreeing with my past self :D
<plowsof> lol
<rbrunner7> "scanning by 9x" What do you mean here?
<ofrnxmr> I mean currently walleta are inefficient, and scan data they provabkt dont own
<ofrnxmr> And download data they dont need
<ofrnxmr> so my "scandata" series adds new endpoints instead of using getblocks.bin since we only need 1/9th of the data
<rbrunner7> I wonder a bit why we don't have that yet. But I guess we will soon be able to see.
<rucknium> IIRC, nodes already served pruned txs to wallets. What else is there to leave behind?
<ofrnxmr> I wuit on master version, because fcmp changes too much
<rbrunner7> I mean, why nobody so far saw that angle of optimization
<ofrnxmr> And mt master version wasnt compatible. So i built for fcmp. Which also skips tree building when using trusted nodes
<plowsof> there is also varying sizes of local cache that provide varying levels of privacy to decrease historic wallet sync time - lets say 7GB~ cache can sync a wallet in less then a few seconds without any calls to a node (see sech1 closed source gpu powered scanner) but you dont need a gpu to bring scan times right down with a cache. this is something
<plowsof> people have noted would be better suited for monero-lws 
<selsta> rbrunner7: it requires storing a historical cache on the wallet side which not everyone wants to download
<ofrnxmr> plowsof: i also did this with scandata
<ofrnxmr> Added a flag to cache the downkoads for reuse by other wallers on the host
<ofrnxmr> The inefficiency there, is that the node doesnt serve "prepared" data. it serves ingredients, and the wallet has ti cook it. When caching the data, its prepared data, so other consumers dont have to re-prep it
<rbrunner7> Hmm. And if I have a single wallet on my computer I won't profit that much?
<ofrnxmr> scandata will profit, but not from cache
<rbrunner7> That could be the case for the majority of users
<ofrnxmr> Which is why its good for lws
<ofrnxmr> Lws can reuse the cache data instead of asking node for the same data over and over
<rbrunner7> I see, different use case
<rbrunner7> Anyway, nice to see that people don't stop to explore new and alternative routes :)
<sneedlewoods> +1
<ofrnxmr> sech, dh, plowsof, and myself all (i guess) started working on similar thi;s and came to similar conclusions
<ofrnxmr> Things*
<rbrunner7> I am sure that @ofrnxmr:monero.social will get our attention if they go public with that stuff
<ofrnxmr> sech and dh pre-prep the data and put it in a gpu ram to get 8s sync speed
<ofrnxmr> I didnt use gpus at all, but created an endpoint that only served pre-prepped data
<rbrunner7> Ok. Do we have to discuss some further subject still today?
<vtnerd> other than sounds like I need to add prepping to lws, no
<vtnerd> although it may have to be opt-in given the size of the data
<sneedlewoods> +1
<ofrnxmr> @vtnerd: Its about 15gb on disk for the full chain of data atm
<rbrunner7> Yeah, and maybe we will have some wildly differing behaviors of nodes in the future, with people sometimes scratching their head if something is unexpectedly slow ...
<ofrnxmr> Maybe a bit less
<ofrnxmr> @rbrunner7: For mine, at least currently, it only withs with --trusted-daemon which pulls from the scandata endpoints.
<ofrnxmr> when using wallet-rpc locally with.. some flag set (dont remember), it saves the data to a cache folder
<rbrunner7> I think I can close the meeting proper while people continue to discuss. Thanks everybody for atttending, read you again next week!
<ofrnxmr> Thanks rb
<sneedlewoods> thanks everyone
<thomasbuilds> +1
<ofrnxmr> But yeah, vtnerd, you dont need new endpoints to have lws pre-parse the data and create a separate cache
