For other readers, if you follow the link they do say why they didn’t do this in-house from the beginning, basically crummy documentation about how to actually do it.

For the poster, I’d say that yes, if you have the resources to do it, this sounds like something you should do in-house. Maybe use Kubo for NS even if you use Helios for everything else? Kubo might be more mature and easier to set up for something like this.

Also, the main alternative to IPNS is DNSLink, and I don’t see that mentioned in your blog post. Maybe DNSLink would work for your case, and it sounds like you’d have the centralized platform to operate the nameserver needed for it, in-house.

The documentation doesn’t describe a sustainable funding model needed to actually fund this offering perpetually.

If I pay a one time fee, what happens when that fee has been used up to pay for storage services?

Larger files get split up into numerous small blocks, so it can cause issues when the requester has to track down and request all of those separate blocks.

I’ve never seen anything saying IPFS was breaking compatibility with the old library, so it should be fine.

I assume you’ve seen the Helia git pages? It looks like they have a bunch of examples, though I’ve never tried to implement them.

https://github.com/ipfs-examples/helia-examples

Well at the risk of being unhelpful ( :) ) I’d take a second to reevaluate whether I really need the huge files in the first place, or whether it would be better/possible to have the content of the file unpacked natively inside IPFS.

IPFS is just not really optimized for big binary files, and you’re running into that. It has a ton of features for collecting and connecting atoms of raw content outside of files, though, and if your application involved content that could be handled natively like that you might find some of those features to be a helpful bonus.

Think of IPFS as a database, not a filesystem. Using it for huge files is akin to putting the file in the field of an SQL table. It’s kind of awkward.

Anyway, I also worry about performance when people start talking about big files. That comes with A LOT of overhead. However, I have heard some people talking about getting acceptable real world performance.

Looks like some good work on IPNS to improve caching and efficiency of those lookups.