Crypto mentality needs to shift to running and developing software that runs locally in addition to providers. That’s how decentralization happens. 

There’s too many lazy people who want everything to be free.

DX Engineer | Head of Developer Relations at /pinata | Building orbiter.host | stevedylan.dev/links

based

I think the main blocker to things being more local-first is generally UX

since there's less monetary incentive to build such products, they are usually made by enthusiasts and don't always fit to normies

building Stauro ❇ frontend dev at @ensdomains ❇
prev. @rainbow, @geniexyz, @fleek

100%, and we've seen this first hand with IPFS. Of course there are levels to this as well. Some protocols are easier to run light clients than others, so there has to be a lot of intentionality with design. Perhaps the answer is building something so good even our own crypto communities can't help but use it.

IPFS suffers from many problems to be fair

1) poor performance
2) high understanding barrier (it took me around ~1.5 years to fully understand the protocol, how pinning and GC actually works etc)
3) decreasing node count, leading to 1 (and centralization)
4) assumption that everyone is going to install a desktop app (not gonna happen)

good thing though that now IPFS folks are focusing on running a node right in a browser (Helia), which should help with decentralization and performance without users even thinking about it

All pretty true, although 1 is generally from poor optimizations on the client side. Depending how you set things up you can get pretty good performance. For instance I have an IPFS node on my Raspberry Pi, and I have it peered with Pinata's infra so I can access stuff on a peer to peer level faster. When people setup IPFS Desktop they're not gonna have that out of the box, and in general IPFS Desktop is a HORRIBLE client lol. Hence why I think we need to go back to the drawing board on some of this.

On that first point I think that really depends on the situation. In the case of an individual hosting they could either be behind a firewall, or they could have settings that hinder their DHT announcing. There are settings you can enable (accelerated DHT for example in kubo) that make announcing better. Peering with nodes and other infra like Pinata also helps. 

Same points above apply to gateways

It depends on the provider but I would argue this is generally untrue. Popular IPFS providers in general have very good performance and uptime. New implementations like Elastic IPFS and Rainbow have helped a lot, but of course the best case scenario is someone using multiple providers and their own node. We actually will be releasing some new OSS that helps on this front. 

Very true, deployment needs to be easier

with performance I meant for example how much time it would take to discover a file on the network, sometimes it takes literally hours

and IPFS HTTP gateways tend to be very slow with that as well

another problem is that IPFS providers don't replicate across diverse infra, so it's bound to only one server/cluster, that's also why perf often sucks

I wish deploying an IPFS node would be much easier than it is now, thus making replication easier