I don’t follow calle, but thanks for proving my point. cya
nostr:npub1xtscya34g58tk0z605fvr788k263gsu6cy9x0mhnm87echrgufzsevkk5s don’t you run a notification server that crawls the network and also one of the largest relays.
I think we need either extensions to relays or a new kind of aggregate data server in Nostr plus some changes to the nips to make changes and sets of data, etc….
i don't really need to run those, but i still think its smart to run at least one relay for your users just so you can do notifications
I love this debate, seeing people who are pro knots has been a great timeline cleanse. I feel like my feed is less dumb now.
what do you mean? nostr:nprofile1qqsraldwhvwcjgltmxwfu7kw8dqef2692yhzheuurd7k3kfy8cxjdqgpz3mhxue69uhhyetvv9ujuerpd46hxtnfduq32amnwvaz7tm9v3jkutnwdaehgu3wd3skueq93uc7e doesn’t run any special server stuff to my knowledge, just plain relays.
he is saying that is the problem, i still think its doable without it. we already have reaction stats in nostrdb and will have them in damus android/notedeck soon
I'm seeing about ~27%-32% of core nodes are cloud nodes based on the 1450 core nodes I have sampled
i can't vet anyone elses data, all I can say is what my node has seen over the course of a day
at least the ones I connected to
no clearnet knots nodes are cloud based, its all residential
looks like core nodes fit a power law curve, where there are many cloud nodes, but there is more in the long tail
cloud nodes: <500
other: >1000
this is not the total view of the network but what I've been able to collect over a day or so (fresh data!)
nostr:note169raxzhad7tejz0eam09w224vkem3usnnusrq7rlnjpxhmq2zxsq535gxj 
looks like the top cloud ASNs don't move the needle too much due to the long tail. but its not nothing
note:note1cyunel733r78nx39jc6grcxwjl2ycsgw9e6zvjn3xskcf5z8k8ks0lt4av
actually the long tail wins here, cutting out cloud nodes bumps knots ratio from 8.2 to 11%
this was a claim by someone on twitter, i set out to validate the claim. people said I was spreading this "lie" when all I said was I was starting to look into it. pretty incredible.
“(..) only seeing 8.2% of knots nodes on clearnet”
Why not include all those who run Knots on Tor-only StartOS from nostr:nprofile1qqs9df4h2deu3aae83fmet5xmrlm4w5l9gdnsy3q2n7dklem7ezmwfcpzamhxue69uhhyetvv9ujumn0wd68ytnfdenx7tcpzamhxue69uhhyetvv9ujumn0wd68ytnzv9hxgtc5fhmq4, Will?
i already said I would do this, but i'm doing an asmap analysis atm (see where they are running, cloud vs residential isp)
yes I suspect you are right, I can calculate that one sec
there seems to be a large number of cloud core nodes. most of the knots nodes appear to be running in residential ISPs, which makes sense. there's just not a lot of them compared to bitcoin cloud infra. I didn't realize there were so many cloud nodes but I guess that makes sense with all the services out there.
AS bucketing is really cool, it turns an ocean of IPs into something readable
https://cdn.jb55.com/s/0eff8dd228e40fa6.txt
nostr:note1s56tp6kx8snvufnnphg3rm5ttg2d5jc0f4pwsuvagqcxv4vf4l9qjaaj3p
yeah for sure, just wanted a baseline comparison since tor introduces a lot of potential funny business
after running this for a day only seeing 8.2% of knots nodes on clearnet. this isn't even AS bucketed yet.
1540 unique core ipv4 peers
126 unique knots ipv4 addrs
this is without tor to avoid tor sybils (creating multiple tor endpoints per node)
I wonder if someone really is inflating their numbers via tor tricks... hmm. will continue the investigation in a couple days to gather more data + tor stats.
nostr:note1mx5pgmksml39elvh5u20pfl5n3z5t3m7lk6xtrk2kscy64hdpvysumtd8m
also deprecated doesn’t mean guaranteed to be removed. It can be used as a discouragement of use
also not true. this is the comprehensive answer that knots people have yet to reply to because they can't
core didn’t even remove the option and there is well reasoned justification for the change. what is your response to cores most comprehensive answer to this debate?
Or are you now stuck with personal attacks now that your narrative has been completely unravelled.
at no point did I say it was 100% definitive. it was an interesting observation. it is mainly because most knots are tor nodes, so there's no way to detect a sybil through that.
so turning off tor biases the data a lot, but allows us to detect potential sybil attacks when comparing asmap on/off.
right, so when I collect data from peers on my node, aws nodes wouldn't be included in the stats. the point is to get a better sample of real people running nodes.
yikes this does not run well on phones
mute packs? xD
my notedeck has more than just my contact list. It follows hashtags, as well as loads stuff from threads.
This is from people I follow, i don’t follow spammy people
lol never gonna move away from that.
why would it be messy?
because uncapped op_return is the actual rules of the network. capped op_return is an unenforceable policy that is easy to route around.
now if we think this is bad, we should softfork an actual limit so its impossible to get around.
I have said this to knots people but the fundamental disagreement is that they believe filters work, even if its not a perfect solution. my take is that the only reason it seems like its not working is because there isn't a lot of incentive to use larger op_returns except for one off trolls.
there are protocols like OTS that use it frequently, which is why you see 99% <80 byte op returns. if there was a protocol that used >80 bytes then i'm sure we'd see that change.
it's not because of the relay policy (imo), its because sha256 merkle root hashes for things are usually all you need for these kinds of protocols.
another thing I was too lazy to implement and now is already getting deprecated. love when this happens lol
I use apple for the same reason, i just wish they weren’t such a pain in the ass to make apps for
llms for osint is fun
https://chatgpt.com/share/68c0beda-62ac-800f-bb18-642efe9d5111
