Paying a relay is the most basic requirement for it to become incentive aligned other than controlling who can use it "for free". My vision is that people will offer relays that charge per use. These relays would have to measure use and as storing an event begs the question, for how long it will be stored, it's obviously not trivial to put a number to each use of a relay but I suspect that the long term most viable relays will be those that offer the lowest price for the highest value and my hope is that it will be a commodity and not just a bunch of premium relays nobody understands how to compare.
Discussion
But your vision supports different relays having different ways of discriminating users? Or do all relays should just work by charging readers? Do they also charge for publishing? Can they have different prices, different currencies? Do they charge per public key by having some credits in an account or do they require each connection to pay for its own usage?
Either you have to answer all of these or you have to allow for relays to have different policies.
Also if I want people to read what I write why should I use a relay that will charge them? Maybe I want them to read for free, now what?
And if relays can't discriminate between who can write and what they can write how do they prevent being spammed? By charging higher prices? Then if someone with a lot of money is spamming a lot that means everybody has to pay more?
I "allow" for other policies but I don't want to support them. I want to support the commoditization of nostr relays and relays that cannot easily be cloned are a concern to me. If people use relay X because people use relay X, there is a problem. We have to get away from this kind of server-centric network effect. "Just use relay X and people will see your posts" would be very bad for nostr.
To your other questions, the exact design of actual pay per use is very under-explored but I would pay "my" relay to serve my notes to my followers and relay operators would probably design some fair-use rules that wouldn't prevent people from reading recent posts but in a future where all are used to automatically paying mSats as they browse the web, they probably wouldn't mind paying the equivalent of $0.002 to load their timeline from such relays.
I don't see how spam would be related. For me, nostr is censorship resistant, so ideally it allows all to broadcast whatever, so the recipient already has to filter and the relay is only charging for the infrastructure, not for sorting out what's spam and what not. Currently, curation is mixed with relays but it should be split to reduce centralization risks.
I see. I kinda agree with you on a big part, but I still think full commoditization is both impossible and undesirable.
The reply spam stuff is just one example where relays are required to do something.
Imagine Nostr isn't this messy half-assed relay-based thing, but instead it's a perfect p2p system where everybody is always online and easily reachable. There are no relays. I imagine this is close to a system you would like to have, ideally.
In that system spam is free and everybody has to run powerful and imperfect antispam systems that filter a neverending stream of spam.
Now in the current world in which people run Nostr on their phones that becomes even more unfeasible. It gets worse when you realize that relays have to be chosen to be read from, and, worse, relays have limits -- so if you ask for the last 500 events you might get 500 spam events, and even if you filter all of them locally you still only got those, and you've missed all the important events you actually wanted to read.
Does that make sense? Do you see my point?
First of all, spam wouldn't be free. I would sponsor my follows and those I reply to to get my messages but not random spammers to get messages to me. Who would pay for the spam? You say the spammer might just have more money but he'd not have to pay merely slightly more than legitimate users as most clients would filter it out and it wouldn't be the purpose of paying for 500 events to just eclipse me. Great, now they eclipsed me but I still didn't read their spam. And I paid to get those 500 events? No, I did not as they were not by my follows, so my budget is still fine and I request the next 500 or ideally the next 500 from people I actually care about and only after that, what others pay to get to me.
That said, I'm not as worried about the filtering relays as I was earlier. I'm just worried about concentration and losing the control at the periphery again. With relays sharing all events I see no way of lock-ins.
Why wouldn't spam be free?
Assuming the world is a perfect "broadcast-for-all" then the spammer could host their own relay and pay himself to spam, i.e. it would be free.
But of course in practice that would be meaningless since you wouldn't be reading from that relay, no one would, because relays are different, they are not commodities.You agree with me on this, right?
It's easy to not download spam from people you don't follow in your feed. But how do you deal with people mentioning you, replying to your notes, or replying to notes from people you follow or just happened to click?
There is much to explore in detail but to answer your questions: If they mention me, client could check if they have followers in that thread or if my follows follow them. This should be cheap to do relay wise as the client could load the follows list with each profile it sees. More involved checks would be replies in the last hour, age of the account etc. but I would like to get these posts not hidden but marked as potential spam at a certain threshold and hidden at another. This way, new accounts still get a chance to get discovered but I'm very conservative when reading notes from people with zero followers I know. Bots will just drown us with engaging content and no relay can protect you from that.