IPFS JS Core Dev Team Weekly Sync, 10 Jun 2019

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: ⚡️ⒿⓈ Core Dev Team Weekly Sync 🙌🏽 June 10, 2019

Description

https://github.com/ipfs/team-mgmt/issues/650#issuecomment-500421092

A

Alright, hello, everybody and welcome to the IPF SJS core team, weekly, sync yay, a lot of holidays, so a lot of people are out, but we will go ahead and get started since we are so short, I'm just gonna go ahead and take note taker, and we will jump into this. If you haven't already added your name to the attendees list, please do so now and also add your update. If you haven't done so and we'll jump in with Jim who is out, she was working on ipfs camp course seed.

A

It's looking at some test, lab stuff and he's out this week on vacation and is gonna, be working on finishing up some ipfs camp stuff. If you have any questions for him, you can email him or wait until he back all right next up me yeah. So last week and this week a lot of prep for ipfs camp course B as part of that I was running through some annoying things with Lib p2p like let PDP creation, so let P be now.

A

Exports create let p2p method the link to the docs are in the crib pad. Basically, that lets you create a Lapita p instance the same way as the constructor, but you no longer have to create a peer info in advance. So if you don't care about the peer info and the default works for you, you can just create that and move on with your life also fix an issue with the switch where, when we blacklist, if an incoming dial happens, will reset the blacklist, we were doing that previously.

A

So this helps especially on local if you're booting up one node before the other, and you do mdns discovery like you- can end up breaking some connection there and be blacklisted for a little bit so that will get corrected yeah and then, along with my PFS course, be work. I'm going to be working on connection prioritization and limb p2p. So what that will?

A

Let us do is for things like jeaious ipfs in the browser will be able to specify our priority nodes like the preload nodes and say that hey, we don't want to lose connections to these. So when the connection manager calls connections, we will calling those ones and then we'll try to maintain connections to those nodes, and that is it for me. Any questions.

B

Configuration is sounds really cool. Is there any update on the delegated routing stuff, like? Is that something that will be ready by a previous camp? We're.

A

Trying to do that I'm not sure yet bashko has been picking that up since I've been pretty busy, but we synched with Michael, Burns and so he's working on getting that configuration so we're just trying to sort through what exactly needs to be configured and how that needs to be configured with go so I. Think we'll.

B

A

Try to get that going before and.

B

Really be like special nodes, or will we use like we.

A

Talked about just using the preload notes for that I.

B

Guess it makes sense.

B

Well, yeah I guess it makes sense, but at the same time like having kind of like a failure mode where like, if pretty well doesn't worry, you can still do I get it Robin because you can still use circuit relay can be a good thing, so so yeah do whatever it's possible. I know. My time is short, but I just have in mind. They're like what's separating them, might be like a wiser choice in there no term right.

A

Yeah and I think yeah having separate nodes but I think at least in the interim that will go, is kind of a work around for ipfs cap stuff and then as because I think. Ideally, we would have dedicated notes for that or we just wouldn't need it, because we have throw a connection but yeah.

C

Okay, any questions I had.

D

A question about the connection prioritization that you were using tags or something like that is all right. Yeah.

A

So one method we had talked about relating it to tagging.

D

A

D

A lot of things that we.

A

Need to new you a lot of things that we, some of the things that we need to do are for, like the the subsystems like the DHT and pops up, and things like that is what we're looking at with connection management is to be able to specify, like subsystems that that potentially comes from a lot of this is going to be worked out in the connection manager, v2 spec, which is still in very, very early draft stage, but being able to tag the connection with all of those components.

A

But we may not. I probably won't do tagging right now, just because the connection manager v2 is going to come out. So it was probably going to be a more crude version of that. But I haven't quite decided. What that's going to look like, but I'll probably put together a proposal this week when I start working on that. So.

D

Beyond connections are sort of like have to be maintained and they'll bring themselves back up if they go down. What other kind of things would there be.

A

So there's bringing them back up and then connection manager, not cooling them. So when we hit the connection manager max, they all get a pass when it goes into the check nodes if they happen to be ranked low, which is probably not going to happen, because they'll probably have a lot of traffic going through them. If they are wrinkly, though we'll we'll skip calling those.

A

All right next is Vasco, who is also out he's working with Michael burns and infra on getting the delegated routing notes up and running, also working on IP NS, / pub/sub Interop issues to do he's.

A

Looking at rhe provider worked for Jas ipfs, because one of the things we need with the DHT is to actually reap revise, so content doesn't expire or our provide calls don't expire after one day, because that's I believe one day is the default or geus or for that th to you provides reviewing a lot of async migration work from kuma vez on DHT, thanks again Kumar vez for that and he's working on a presentation that he'll be presenting with rock at the Avera university for master's work and then a few other things interrupt chess for a hybrid pub/sub network.

A

So I think that's with getting gossip sub working and then falling back to floods up we're going to interrupt for that. Have questions for bashko reach out to him. He should be back tomorrow and I slept Dirk yeah.

D

So I just been working on figuration profiles, which is like you know. If you have a if you're running your nodes in a data center, you can quickly specify the configuration for it and next time I'm going to work on on looking for the pinning system at the moment, there's no locking. So it could cause some problems if people are doing a lot of pin updates.

D

At the same time, I'm also gonna be doing a refactor of the pinning codes so that it works a little bit more like the go system wherein you can do a bunch of pin changes and then flush them all out at once.

A

Cool key questions for Dirk, you No all right, Alex,.

E

Okay, so yeah, so that's what you've got the PR merge for having the IP of us promotion bow to async/await, which is amazing that got released. I integrated hugo's was an implementation of the wrapping chunker into u-. S, which is great also has some interrupts s4.

E

So we can now guarantee that the representations, the same tango chairs fixed all the bugs in NPM on I, give s. At least you know closed with the issues and and working on the road map for package managers. Special interest group next is my pair first cam I'll give us, can Nike first camp, so I own MP, a little box which is going to run the registry that we have running on chair, stop packages or whatever is ready, so yeah registry chair type, your best.

E

Ok, it's going to run on this Raspberry Pi as and him in the box, so I've started. Adding support for mdns can broadcast for the network. Hey.

D

E

Then you can use Antioch EFS, which would then find up and use that, as the you know, make a connection to that directly because there's other packages which would be super cool when the internet goes down at the camp, which of course it's going to happen. It's not gonna happen, but of course it's gonna. Yes, then I going to because so needs kind of gather running so they'd start syncing NPM. So that's so I don't have its own hotel room long time buzzing away like like I did in Cosco anyway. So that is gonna.

E

Be me too okay, maybe tomorrow morning and.

F

E

That the course the p2p course has.

D

B

E

In any questions.

E

A

Questions I'm glad that I am p.m. is it's bug free bugs.

E

B

Is like not that question and just like a curiosity like I, remember like a one of the big pitches, my car dark, because, like the lighting situation here, one of the beaches, 11:00 p.m. on ipfs back in 2015/16, was really like.

B

The install modules without internet, like in cell modules, without collecting twenty place, and I since then like that messaging has changed to oh we're, like the core of the registry, and you can now like just point to this registry when you think about that like voir, is the reason why the change in in that communication?

B

And why aren't we making more like from facing, like the that demo, where, like one person, can like just take that the cable, the error at cables from the the the switch for the router to make sure like there is no internet and still install modules from machine to machine, be I.

E

Think the irony I think it wasn't a conscious change. I think it was just like this is the thing that we've built. Let's talk about it and then people start talking about that anonymous done. I think the thing that kind of interests me is like publishing, I pianist names for all the modules and because then you actually need a centralized registry. A tool. If you can resolve and I pianist named the results to a parking I mean that way. You'd see the funds talked about yeah.

B

But you just be the one I'd be nesting for an entire registry. Right, like you, don't need twelve from.

E

That, because, if you like, if I write, a module and I published might be nesting them cool down that module without needing a centralized anything. Where is if you'll publishing one up in listening for our registry, then they message was yeah.

B

Yeah, like what I'm thinking is like from where we are today today like to just saying that, because I penis will be publishing one record from going where we are today to adding distributed publishing where I people can I own their own publishing to the IFS Network, yes I careless, who have their own agonist keys and and like which their names yeah.

B

Absolutely mmm that's like it's something to consider right like it is definitely important to just like spark the magic in people's eyes like to to remember that I, hey, like here's the thing like when you are offline, the thing doesn't work and B I'm, just crashes or like your neighbor, cannot install the thing, but we aren't able at BFS like that.

B

That's a feature there's already bacon, like it's already like a a promise that is fulfilled in and it's there and I think either the registry yesterday, oh I kind of like, although it provides a Casio piece of infrastructure. It also as hidden like one of these very cool features like I mean I'm. Just highlighting that because, like from all the conversations I've been having like people, I don't seem to get that the thing about empty. My BFS is actually that you can I just read packages from your machine to the next machine and so on.

B

They always think it's just about like having like mirrors. That's it's behind HTTP 10 points but, like you still are like using HTTP anyway, agree.

E

I mean I think.

F

The mirror is like.

E

The least interesting thing the whole thing exactly is this: like yeah, we like entropic, coming along people, Stein's talk about distributed registries and with you know like get Homes registry as well, and all that like now is the time to start saying: hey, look this stuff can be distributed to you know, and here is a toaster for you to do it interesting yeah.

B

Indefinitely, like entropic, it's kind of like a great first step towards that conversation, because, like it's about Federation and information, is not just about like connecting a bunch of leaders in in making them like shared alone.

B

It does talk about like what is the the authority that keeps publishing, like the update to the packages like who says like like who confirms that, like it's, Alex publishing a package right and then like it gets uh psych into the right direction to other conversation or like we can take a step further and I like composing, distributed still federated like the Federation, see important, because you still want to have some authorities that I can protect the users and can provide convenience.

B

But if you want to go like full-length and like just like me or your own publisher, you can yeah. Is anyone like from in traffic coming to people's camp I.

E

Don't think so, cuz the thing you've been around for I come on well.

B

Is it worth still inviting them I think it is yeah.

E

Sure absolutely.

B

When I shoot an email.

E

A

All right do we have anything else. There's nobody here for crusty updates. Other notes. Vasco is out on Friday for his presentation and then I am technically out next week, but I will likely be around working on ipfs camp stuff, but yeah.

C

If that is it, I'm.

E

Sorry one thing a lot only like inviting like Chris, Dickinson and CJ sir I. Seen that mean it's your of us can do you have any rooms left?

E

B

Check with events theme like that response, I've been getting is like if it's super important for the success of the project or for like whatever conversations we want to have, they can make a way to figure it out, but it requires some shuffling. It would be great if, like for the events team, it would be great if we have thought about this like a month ago, but I can so your.

A

Team will sleep outside and tends to make it.

F

B

Well, you can always I sleep by the pool and I.

F

B

Yeah but I I think it's doable if it's important.

B

All right thank.

C

You all yeah, that's it all right! Thanks! Everybody see you next week, bye.