Radicle Architecture | Hacker News Reader

Radicle Architecture

radicle.xyz

71 pointsjkarni18 comments

When would I want to use this? What problem that people currently have does this solve? The website tells me what Radicle is and what you can do with it, but as far as I can tell it doesn’t explain why/when I would want to do that.

When a blog about a collaboration tool devolves into a formal state machine definition and blocks of lisp syntax, I have my doubts about the project's ability to communicate with potential users.

If you’re not interested in how things work under the hood, there are nice high level docs for you here http://radicle.xyz/docs/index.html :)

Your blog's users can be a technical audience.

Following the introduction tutorial, in which you add to the Radicle Garden[2] (which is a nice idea btw.), I get a 'The machine owner appears to be offline' when I try to propose my changes.

I think the very idea of Radicle is nice, but having a rock solid VCS is crucial to a developers workflow, so they really need to iron things like that out.

1) http://www.radicle.xyz/docs/index.html#installation-setup

2) http://www.radicle.xyz/garden/index.html

After being optimistic enough about IPFS to create a whole service around it (https://www.eternum.io), and mostly because of the service, I now have my doubts. Nodes can routinely not discover each other. I'll add a file to my node then request it from a gateway and it will take multiple minutes before the file is resolved. There's no information on how far along a pin is. The API for pinning things is atrocious (you need to keep the http request alive for the multiple days it might take to pin something). I have to resort to just retrying the pins over and over until they succeed.

I'm sure there's some information hidden somewhere if I shell out to the binary and parse the response well enough, but if a paid service required me to do this, I'd laugh and never touch them again. These days, whenever I see that something is "based on IPFS", I mentally translate it to "not working".

> I'll add a file to my node then request it from a gateway and it will take multiple minutes before the file is resolved.

Sadly, I've noticed this myself. Have you reported these things to the IPFS developers?

> you need to keep the http request alive for the multiple days it might take to pin something

I didn't understand this part. You make the HTTP request towards your local IPFS daemon? Why does it take several days to pin something?

Yes, I've reported everything, and they were extremely helpful (to the point where they debugged things with me), but nothing changed.

> Why does it take several days to pin something?

The daemon doesn't act like a torrent client, where you can add a pin and then come back later to check progress. You need to do an HTTP POST (or whatever verb) and then wait until it's done, which due to availability issues and size might take days to complete.

> Yes, I've reported everything, and they were extremely helpful (to the point where they debugged things with me), but nothing changed.

Do you remember how long ago this was? I'm still hoping this is a solvable problem since I really want IPFS to work.

> The daemon doesn't act like a torrent client, where you can add a pin and then come back later to check progress. You need to do an HTTP POST (or whatever verb) and then wait until it's done, which due to availability issues and size might take days to complete.

Oh. That does sound awful. I wonder why they designed it like that.

> Do you remember how long ago this was? I'm still hoping this is a solvable problem since I really want IPFS to work.

This was around two years ago. I hope it'll get solved too, but it doesn't seem to have been solved yet.

> I wonder why they designed it like that.

I guess it was just easier, but they haven't done anything in years. The ticket has been open since 2016: https://github.com/ipfs/go-ipfs/issues/3054

I recently looked into this, and now there at least seems to be a --progress option to 'pin add' which shows number of "nodes" fetched/processed (IPLD nodes?). Not sure how to put that into greater context though, i.e. how many nodes are there, and how does it translate into bytes? But it's better than nothing.

Ah, I'll check to see if that's in the API, although it's not super helpful if I still need to keep the connection open. Still, as you say, better than nothing!

How does this compare to scuttlebutt and ssb-git? I know IPFS and ssb are different but conceptually in how data is segmented and distributed?

git-ssb was one of the inspirations for Radicle, we're big fans of the project and we've spoken to cel, the author of that project a few times. one of the most meaningful distinctions is visibility in the network, git-ssb repos are distributed through the Scuttlebutt social graph, while Radicle projects are seesed to the whole network. We've written a little on this in the FAQ, 4th question: http://radicle.xyz/docs/#faq

RTFM is an underrated comment, thanks for reminding me so politely. Will definitely review now!

Love the style on this website, in particular the use of Vulf Mono!

Is it just me or does this blog not have links for each entry? There seem to be two entries at present but no way to link directly to either of them?

Does it have a viniculum?

Reply on news.ycombinator.com