Ask HN: What are some of the gripes you have with IPFS? | Hacker News Reader

Ask HN: What are some of the gripes you have with IPFS?

I've seen a lot of posts come up where IPFS is mentioned and someone chimes in and says they don't like the choices they made or how the network is run, but I haven't seen many people go into great detail. I'm curious what these problems are and why they're important.

I'm really interested in using IPFS for one of my projects and all I really know about it are the good things.

View on HN

IPFS person here. I can relate some frustrations / concerns that people sometimes have with IPFS - and happy to also chat about your specific use case and whether IPFS is a good fit!

1. network performance / reliability / scalability The IPFS network grew 30x in 2019, and because of that we've had some performance challenges - especially around resolving niche content via the DHT / handling NATs. We're working on fixes and improvements here, but for applications that require extremely fast response times there are some work-arounds (like directly connecting to known relay nodes / swarm connecting your peers) to get around these issues.

2. access controls / reader-writer privacy Right now access controls live a layer above IPFS - and IPFS doesn't support any out of the box encryption / privacy preservation (coming from it's origins helping replicate public datasets like wikipedia and scientific data). There are some groups like Peergos, Textile, and OpenBazaar which have invested in building these controls and privacy preserving approaches (like a tor transport for IPFS) - but we'd still recommend folks to be very careful when using IPFS (especially the public DHT) for highly sensitive tasks.

3. extremely large data One of our foci with our "package manager" goal for this year is to make it fast to add TBs of data to IPFS. Right now, depending on folder structure that can take a long time (hours) =/

4. "out of the box" website support While there are lots of simple tutorials for putting a static site on IPFS, many website-building frameworks don't write relative links in a way that meshes with "anyone can provide this content" (aka relative urls get wonky). Hugo is a notable exception, and we're working on creating tooling to support other frameworks like vuepress too. IPNS being too slow to resolve is also a complaint for sites that want to regularly update to a newer hash. We have dual track improvements on this - both a new version of IPNS over pubsub that is less reliant on slow DHT lookups, and making DHT lookups faster in general for all requests - but the other workaround is to update dns directly using dnslink - which seems to work for most folks.

Check out https://awesome.ipfs.io/ for the sorts of projects that people are building on IPFS today!

Thanks for the great recap on the state of IPFS. I wasn't aware of continued progress on these fronts and it seems much closer to being generally useful than I would have guessed. In particular (2) & (4) getting better but not realizing that adoption was creating it's own issue (1).

Reply on news.ycombinator.com