257 Members
Implementing ipfs in rust based on rust-libp2p https://github.com/rs-ipfs/rust-ipfs https://github.com/ipfs-rust/ipfs-embed38 Servers

Load older messages

1 Dec 2020
@dan.shields:matrix.parity.ioDan Shields joined the room.00:01:23
@rklaehn:matrix.orgrklaehnNot sure if everybody is aware, but it is the time of the year again where Protocol Labs is asking the community for proposals what to focus on next year.08:54:30
@rklaehn:matrix.orgrklaehnI sound like a broken record, but I made a proposal for scaling the DHT... https://github.com/ipfs/roadmap/issues/7608:55:06
@dvc94ch:matrix.orgdvc94chmedian lookup latency 200ms in a dht with 9.5mio nodes, by optimizing only the local implementation10:47:47
@dvc94ch:matrix.orgdvc94chthis is what inspired the current bitswap design and custom syncers to some degree, building the lightest server and offload all the work to the client10:49:09
@dvc94ch:matrix.orgdvc94chalthough I must admit to only have read the abstract of that paper10:50:07
@rklaehn:matrix.orgrklaehnOne of the tracks of my proposal would be to actually implement all those papers and see if their findings are reproduceable using something like testground.ai .10:54:52
@rklaehn:matrix.orgrklaehnAnother track would be an extremely easy to use large scale performance test kit, basically "put your binary here" instead of having to pay and set up your own large scale testing infrastructure, which is infeasible for small projects like ipfs-embed and rust-ipfs.10:56:24
@rklaehn:matrix.orgrklaehnThen the different ipfs impls could duke it out by the end of next year... That would be fun.10:56:54
@dvc94ch:matrix.orgdvc94chWell we do have netsim embed10:59:51
@dvc94ch:matrix.orgdvc94chThe problem with testground.ai is it requires a http interface11:00:16
@dvc94ch:matrix.orgdvc94chAnd it's written in go deployed using the gx package manager11:00:46
@rklaehn:matrix.orgrklaehnI did not know that. I thought it was language independent despite being written in golang.11:01:15
@dvc94ch:matrix.orgdvc94chAnd that never worked reliably for me, compiling a third party go project always seemed like a hughe pain to me11:01:44
@rklaehn:matrix.orgrklaehnYou could wrap e.g. ipfs-embed with a tiny warp server if the HTTP interface is the only thing that is missing.11:01:51
@dvc94ch:matrix.orgdvc94chWell it needs to somehow tell the nodes what to do11:02:07
@dvc94ch:matrix.orgdvc94chOr maybe I need to look at it again11:02:23
@rklaehn:matrix.orgrklaehnWell, the idea would be that all of that is being handled by protocol labs, and you just have to upload a binary to do a really large scale test with 10000 nodes or something.11:02:41
@rklaehn:matrix.orgrklaehn(With a 100000 FIL reward for the fastest impl by the end of 2021... )11:03:30
@dvc94ch:matrix.orgdvc94chwe'll see what comes out of it, I love the idea11:04:23
@rklaehn:matrix.orgrklaehnProbably a long shot, but worth spending 1h filling out the form.11:05:42
@dvc94ch:matrix.orgdvc94chI'm not sure if the current ipfs api maps well to how ipfs-embed works, depending on which operations they choose11:17:27
@rklaehn:matrix.orgrklaehnThe current ipfs api is properly called v0. There should be a new one before they go to ipfs 1.0.11:21:23
@dvc94ch:matrix.orgdvc94chNot trying to be overly negative, just expressing a few concerns11:24:10
@rklaehn:matrix.orgrklaehnyou are right, it would have to be ensured that the solution is not overly constrained by the current API.11:32:54
@rklaehn:matrix.orgrklaehnProbably best to raise your concerns on the ticket instead of in the chat though...11:33:45
@dvc94ch:matrix.orgdvc94chWell as you know we're still figuring out what the api should look like (@rklahen wrote a new ipfs implementation which ipfs-embed will learn from) so before then I can't really make a constructive suggestion13:49:33
@dvc94ch:matrix.orgdvc94chand all my thoughts on the matter at the time of the ipfs-embed presentation are published in form of a video13:52:48

There are no newer messages yet.

Back to Room List