Randomly assigning relays to new users without any reliability or performance metrics will not end well.
Primal has the benefit of their caching service to mask the problem but those users will suffer on other apps that don’t support it.
Discussion
Totalmente de acuerdo Mazin opino igual, no termina bien
I was already blamed once for using slow relays - while the client didn’t show me anything about those relays in terms of their performance.
How can we assign best performing relays?
Something Ive always wanted to add to damus, but maybe it won’t matter as much since in outbox your peers decide which relays you talk to anyways.
🤔
So much of nostr client performance hinges on good relays and we’re assigning them at random?
Seems like that might not end well. Or am I missing something? Can the outbox model be improved?
Some of us are concerned about it others don’t see any problems 🤷♂️
I know nostr:npub1uac67zc9er54ln0kl6e4qp2y6ta3enfcg7ywnayshvlw9r5w6ehsqq99rx has been working on publishing relay data in nostr events but it’s a challenge that needs multiple data providers.
Reliability is more important than performance but it still needs to answer REQs in a timely fashion.
NIP-66 events are are being published full-scale from 8 monitors on 6 continents as of about . 8 pushing 30066 events with round-trip times and 1 publishing full metadata and 30166 discoverability events.
oof, as of about 3 weeks ago*
What is this website?
No it's not, it's https://github.com/sandwichfarm/myrelay.page
ohh, super awesome
thank you for sharing 💜
There's a live stream of data here: https://nostrpul.se
Yes at one point nostr was unusable until strfry came along. If you get randomly put into a pool of unreliable relays then primal users will effectively rely on the primal caching relay for everything. Afaik all primal users have to use their relay anyways since it is used for pulling stats and stuff. I’m not sure if outbox even makes sense in clients that are centralized like this? It seems like a decentralized-in-name-only kind of setup.
Let us give our monthly thanks to nostr:npub1yxprsscnjw2e6myxz73mmzvnqw5kvzd5ffjya9ecjypc5l0gvgksh8qud4 - whatever he may be up to!
A true legend 🫡
actually damus would probably have been on the outbox model by now if it weren’t for strfry, it carried the relay pool model so far 😅
Its great for reliable user experience … but no so much for decentralization. distributing query load across smaller relays would have been the only other way to do it, but it still would have been slow.
Social clustering. Write to the same relays your peers write to. If you know a bunch of people, odds are if someone knows you they know them too, and they will be able to use the same relay to request multiple peoples' notes. Random distribution is the exact wrong solution @milj
Sorry, I don't mean to overstate my case, this is a step forward for sure and I applaud you guys for prioritizing it
Yeah, I agree with this
yes, and most of the relays are written in C++ or Rust
so, hmmm 🤔
There are so many different "kinds" of relays with different policies that random selection is flat out not going to work. It took me ages to find even 20 that meet the requirements of hosting new users for free (and yes, yes, I will list paid relays with links to their signup pages at some point).
Absolutely, and the research you did to get to that list was well needed.
I’d be surprised if there are even 100 free public relays that are fully functional (and usually online) so selecting from 1100 seems insane.