I'll write a lengthier reply now that I'm on my desktop, too much to finger tap ;)
I want to pre-face that I'm merely adding nuance to the conversation, not making excuses. There should be no excuses for fixing actual bugs, perhaps other than that they are super hard to fix.
First point I want to touch on is the idea of cross-compatibility between clients. I don't think you brought this up, but I know others have in the past. Clients having different features that do not align with one another is a feature, not a bug. Nostr does not have a defined spec beyond what kinds you can use. All the NIPS are possibilities, not requirements. So, with this lens, I think it's important to understand that different clients will have very different features, and that is by design. It's something to embrace, not to try to "fix". Again, I don't think you said anything about this, but just putting it out there for anyone else reading.
Relay management - this boils down to content as I said. It doesn't matter how well you manage your relays, if the content is not there it's not there. Could the UX get better? Absolutely. Will it help? I would guess not at this point. I use Jumble daily and I browse different relays and see seemingly different content but ultimately its the same people saying slightly different things on different relays. It's boring. Because of this I don't even want to check the relays. I do just to see what's up, but nothing is up. To truly make relay management useful and interesting, you need a lot of content to differentiate relays. But, if you think about the kind of garbage you see on X, even their numbers won't warrant super interesting relays if they switched to that model. You might get a reddit-like browsing experience but ultimately it's the same 1% of users who post everything. It could be interesting and useful at scale, but not at our scale.
Pull to refresh - This to me sounds like a feature request - one that warrants active decision to include. As you know we have a different culture here from what we consider "normie fiat" apps. We value other things. For me, one of those things is not hijacking my attention to the maximum and gamifying every action. I know some clients have chose to go down that path (yakihonne) and I even considered it at one point and proposed myself, but I've come to a realization that I would rather build in mechanisms to prevent or limit addiction than to add to it. Pull to refresh is a gambling mechanic that I have considered for x21, but ultimately chose against. This was a conscious choice not to include. But, since we have such a different variety of clients, people can use others that might choose to implement it, or already do -- I don't know if any do because I haven't done a fresh round of tests recently.
Discovery / onboarding - discovery is a function of content again. Yes, we can introduce methods to get people to subscribe to hashtags, to pick interests, even centralize feeds (Primal) or even introduce algos, but ultimately all of those things will boil down to not showing anything interesting to the vast majority of dopamine-jacked brains that expect instant gratification, infinite loads, pull to refresh etc... This is a walled garden / community issue to solve, much of which depends on marketing initiatives, not so much design. We can improve the design but it's a fruitless effort at the moment. Primal already has good onboarding, better than any other social media I've tried (I tested all major ones), and the feed is still full of bitcoin content.
Infinite load issues - these are probably bugs and bugs should be fixed. I will not make excuses for persistent bugs. BUT, I want to give credit to devs who are fixing things all the time and have had github issues outstanding for some time trying to get to all of them. I also want to give credit to people like
@elsat who started interoperability issue tracking and devs who actually try to fix those things, BUT, I would also point to the very first paragraph to say that some of this is not a problem, but a feature, by design.
DMs - I can't speak much on this, I have tried Keychat and it worked fine for me. There were issues with names or something but I have not tested recently. It seems like White Noise might be good too, though i haven't used it much. Client DM implementation is an individual choice. Some clients (like x21) do not even want DMs -- this is developer choice, nothing to do with UX. It's a bitcoiner culture thing, not something that must be hammered to perfection.
Content fetching / loading of reactions etc.. all has to do with relay connections and caching. This is ultimately a difficult problem and requires local solutions such that
@jb55 is working on. Relay on device type of thing. It'll get better. This is a technical / engineering issue, not a designer UX issue.
I write all of this not as an excuse, but to shine a light where others may not have looked or considered. There are limits to what we can replicate from the central server world, and we need to consider how much effort we should make, vs embrace what we're better at.
What bugs me, is people who swing by overnight (not you), and claim to know the problems of nostr (UX) and offer quick solutions (as I tried). It's as if all of these things have not be considered before. Like, we are too dumb to see that other apps work buttery smooth from a database. A lot of these observations are rooted in experiencing ONE client and not even trying the other 90% that do those things that the one lacks in MUCH better. I've tested the onboarding of all fiat apps and they all suck. What makes them sticky is their content. We can focus to improve the 20% roughness of nostr clients, or choose to find ways to bring more people here that will actually stick around -- I think we should focus on finding ways to bring people who will stick around.
On that note, there's a lot to be said about focusing on the strengths of nostr - exploring other stuff and how it can be interwoven with the social stuff. And we're doing this every day. I'm doing it, and I know others are too.
