16:00 "fission drive" sounds exciting, so in principle I can have data on my disk, also distributed via IPFS and available to local-first apps. This is something I've had in mind as a "data mirror" , it would be cool to interoperate with data from various silos and build apps/alternative clients on top of it. I've got a bunch of data extractors/normalisers already , would be interesting to try and connect it!
By the way, for demonstrating offline mode, both Chrome/FF devtools have network throttling (which is also useful for triggering/testing race conditions). In Chrome offline mode is also in devtools, in firefox offline mode is in "menu - more - work offline" (which is a bit less convenient, but still)
Does anyone that has dug that rabbit hole deeper, than I have, has any insight on why doesn't this architecture seems to catch on? I feel like it's a monetization issue, but have nothing to back that up.
>Our guiding principle for using IPFS is: Everything should work in the browser, including on mobile, with no extensions required.
I wonder how these two cope with one another. To me, you're either not local (must connect to be sure your data have persisted) or you need something more than just browser. Otherwise you're at the mercy of browser vendor not to destroy your in-browser data with borked update.
Or do I incorrectly understand "local first"?
> Or do I incorrectly understand "local first"?
I believe that the specific term "local-first" comes from Ink & Switch: https://www.inkandswitch.com/local-first.html
> I wonder how these two cope with one another. To me, you're either not local (must connect to be sure your data have persisted) or you need something more than just browser. Otherwise you're at the mercy of browser vendor not to destroy your in-browser data with borked update.
Maybe a clarifying one-liner would be "it's local-first, not local-only" ;) There is totally a tension there from not controlling the browser itself. You're absolutely right that the browser can clear it cache pretty much whenever. In practice, this appears to happen less than advertised (the extra caution is a good default!), but it totally does happen. It's something that we've spend a bunch of time trying to figure out the balance on. We also have to recognize that we live in an online world, so there's a bunch of additional work going into making things feel instant when connected. Fission is pushing on both sides of the problem to basically get as close to "your data everywhere all the time" as possible.
The principle is that you should be able to show up to a web app and have it "just work" and not know that anything special is happening. Everything should work if you loose connectivity, or are on a plane (if we ever do that again). You should also be able to self-host your data, or have it cached in multiple places (which is a big advantage of IPFS).
Fission provides one way to automatically keep a second encrypted-at-rest copy outside of the browser while you have connectivity. There's nothing special about our IPFS nodes; they're just set up to make that smooth for the kinds of thing that you'd want to do with the SDK, and add some niceties like DNS automation so you're not always fiddling with hashes. Part of us presenting to the SIG was to try to get better access to local desktop IPFS nodes, so that those who want to turn the dial to "fully offline" can go further with that.
So IPFS handles the actual 'local first' aspect in the sense that once your data is cached, you can access it, and once you regain connectivity, it syncs again
If you want to publish something that other people on the internet can see, then yes, you need a path off your machine, such as WebSockets. This can be a peer IPFS node, or to Fission's servers, or your own self-hosted node to provide availability.
Part of presenting to the offline-fist SIG was to get more support for parts of this, like communicating between a desktop node (e.g. go-ipfs) and your in-browser js-ipfs instance, which doesn't work out of the box today.
In the case of IPFS, you can have a IPFS node running locally, have a browser client communicate with it and just store the latest hash of your state (that you can backup outside of the browser too). So if the local state (in the browser) gets lost, you can restore it from the hash. Often used content can be cached in the browser but still stored in your IPFS node.
In my opinion we should:
1. Specify new data link and network standards for decentralized networking over private and decentralized networks
2. Build the networks
3. Start to think about high level stuff like IPFS