IPFS GUI and Browsers Weekly, 12 Feb 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: /dnsaddr/ in js-libp2p, GUI refactors, playwright tests - IPFS GUI and Browsers Weekly 2020-02-12

Description

About IPFS GUI and Browsers Weekly: https://github.com/ipfs/team-mgmt/issues/790
IPFS Mirror: https://ipfs.io/ipfs/bafybeid5jcqj6efkop4qzfj7himkele5h3w3kqtpkxaa6osxhx2ygemnfm/

For more information on IPFS
- visit the project website: https://ipfs.io
- or follow IPFS on Twitter: https://twitter.com/IPFS

Sign up to get IPFS news, including releases, ecosystem updates, and community announcements in your inbox, each Tuesday: http://eepurl.com/gL2Pi5

A

Welcome to GUI and in web browsers, 412 of February 2020.

A

This week we have some items on agenda and for some reason, I'm always first. So maybe let's shuffle this a little bit this week and Jacob, can you go for.

B

Outrageous right, yeah, so I just wanted to briefly talk about DNS over HTTP, so we have need to start. We need to get this implemented because, right now in j/s we don't handle resolution of dns adder multi editors. This creates problem for like our Gateway nodes and anybody else who wants to use DNS and trees and get text record backs. So we're going to be working on implementing resolvers for live p2p. I synced up with hygge earlier.

B

So what I'm gonna do is based off what we talked about in the prior art for JSI PFS for IP and s over Tina's I mean going to work on creating a gist for describing how this will work. But the basic idea is that for JSF p2p you be able to provide a map of resolvers and then those does that resolvers would be configurable and then we'll just start out right now with a small resolver for handling the DNS errors, because that's what we have the highest priority for right.

C

B

That will let us handle resolvable, so then JSF PTP will store, will still store the dns adder associated with the peer and then when we need to do dials or any lookups on the multi adder. We can do the resolution with DNS cache based on the the TTL, so I'm gonna be working on that the end of this week, I get a just a proposal shared out for how that will work and how that will look like for a configuring.

B

Lib be tepee, but the idea is that Libby tepee will also come with defaults for that, so people don't have to worry about creating resolved us for that zone.

A

The question I have had this page is sort of like a background for folks who are not familiar with next a question and clarification. Oh there's an issue with sort of like more details, background in case anyone is interested. It's on the edges, IP of house, but we just like check out said: we need to probably tackle DNS other first in lately to be probably the way it will work.

A

I think when someone will initialize Jessica TFS, we will always pass that configuration to the p2p I mean if someone wants to use custom, DNS resolver, let's say a DNS server is DPS. Had some service provider, probably it would the way it would look in the end. It would look like in Jessica's. It would be like passing that to constructor and then internally a sacrifice would pass that philippi cookies, that's more or less that long.

B

Yeah yeah so we'll be doing everything everything up to where Jess ipfs I look briefly at the issue and I'll make sure to go over that before finalizing proposal. The HTTP resolvers were more specifically towards the the hosts of the resolvers and then for for JSI PFS. It would likely just be instantiated if it needed to override it would be instantiating.

B

The certain resolvers that lit p2p will use because it may in theory, people can provide resolvers for whatever and it might not just be HTTP resolvers, but that will be the thing that we're focus on first, because we have a very specific use cases we need to solve. For that.

A

Yep totally super cool, make sure to see. See me when the proposal is more yeah. Yeah super cool, alright, the next one is what I think a factor. The recent refactor of a p2p and J's ipfs means for our user facing applications, namely a companion web UI hand desktop self assessment mode that multiple times so I feel makes sense to. Could you go over them and mention had all pages pre-loaded before I rebooted, my laptop so from IP. First companion perspective, we use internally, both J's ipfs full note.

A

We also run J's ipfs, HTTP client, so user can pick between external and embedded note and in brave. We also use like we make a more extensive use of embedded JSF us where we provide custom Lippe to the config and modules for things like TCP transport and more like more closely tailored version of embedded JC DFS. For that brief environment, and in all those touch points, we will have to basically a probably do like a factor of IP fat compile code.

A

That's why I did not work on companion for some time, mostly waiting for dust to settle down around ipfs waiting for the next rid of Jessica Fez, and also waiting for signal instruction, because that embedded node is right. Now, relying on WebSocket start clicking brave. So unless we have either that web RTC one or Stardust's ready companion.

C

A

Sort of like a blocked and we are not able to switch to the newer version yep, that's the topic of member. The detective has embraced it's a separate, separate set of challenges, namely with some IP eyes. Chromium ass API are being deprecated by Google, which gives us like a soft window until 2022 to figure it out next steps for brave, but not like separate branch of work on the ITF s web UI, which is our dashboard and interface for, and it's shaped with go IP of us.

A

We chase a DFS with IP, first companion and also with IP felt desktop, so this dashboard is either bundled wave or is open from ipfs by those projects, and it's basically web app in react which connects to ipfs HTTP api over to ipfs api of our HDD client, which also got refactored and we'll meet someone to give us some time to go over. All the users touch points with our api and refactor that to the new api and for a people desktop the surface it looks similar is similar to IDF us web UI.

A

This electron up, however, it does not have as many touch points with idea, a J's ipfs, HTTP client, programmatic interface. It's mostly like just checks, if notice out and for everything else, it opens web UI. However, in all those GUI applications, we are still using the old callbacks and promises based API.

A

Before did this refactor with introduced some breaking changes, so the patience we gets it out eventually, but I feel we want to take it slow, especially given that in ipfs web UI we got end-to-end tests, which I started working here today to make sure they work with both old and new programmatic interface from the J's side of things.

A

So that would enable us to sort of like randos and run tests, as we are transitioning both like web UI, but also go active at Jessup EFS, which also will run those intern tests and handle situation when we have like one part, front-end or back-end, on of those end-to-end tests being using the new ITI, and maybe the demon is l1.

A

So that's more or less the current situation and high-level overview of where we are with those with those factors.

A

I believe that's it from me.

D

I will have seven thing regarding the starters, since you mentioned it, so basically I've been pushing forward with it in the last couple of days and now I got to the the PR finally ready for an initial review.

D

So basically, there are some things to eventually polish regarding the encryption, because the initial version of star list was not using Sakai, oh and now, with the current dollar I'm using it it's working where it's using the secure encryption which we might not want to use, we we are still discussing if it should go with it or not anyway. I hope that we can get it March next week and after getting it emerged.

D

Basically, we also need we'll need to deploy the starless new version of the server and hopefully, after that, the third list will be ready to be used, and that's it. That's.

A

Awesome yeah, we will probably need more than one I would I would feel much better if we had like two or three just two just two from the get-go. When someone learns about Stardust and like signal signal, this temporary solution of signaling servers like in every example and places like that people should see like more than one just to have that at the back of their head. A good habit to have redundancy otherwise I am not sure what will be the default.

A

I believe the current default in, like the default bundle of J's ipfs, is to fail if no like web socket. Star servers are available. That.

B

Won't be the case, I think with the refactor I need to double check, though, but our dialing logic is our listening. Logic is different than it used to be before the refactor. So now we don't just like the addresses we take it we're more selective and we spin up listeners for each address. So we allow listeners to fail, but it could be possible if we need to check, but that's also should be relatively easy to configure if we want to do an update to JSF BTP to allow like okay if WebSockets are listening, is failing.

B

Let's not not fill out right, let's allow that failure, but we may not want to like right now it if all WebSockets our addresses fail. It will fail.

A

In I touched upon that around the works done for brace so embrace I changed because the module for like WebSocket star WebSockets are multi. It enables a mode when you could allow failing of all of them, I believe so I slipped the flag slip, the default for brave. So even if our signaling server is down, the note would still both that's probably like safer default.

A

Otherwise, like people may get surprised that at some time nothing.

C

A

Instead of just like subset of connectivity,.

A

Yep, do you want to show us something.

E

A

Think I feel that you want to show us something you.

E

Want me to show it Jake you've already seen.

E

Okay, let me share my screen.

E

Okay, so I've been working on.

E

Basically, a test runner based on playwrights playwrights is basically the puppeteer 2.0 that supports chromium web capes and Firefox.

E

And I'm gonna I want to run des in browser, but not to use karma because it's awful so I tried to make it work with playwright and turns out. It's super simple and neat allows for some extra nice things that we couldn't or would be hard to do with karma.

E

So you guys are seeing my screen right.

B

E

So simple stuff run the tests some console logs this physically, but knows everything with a with webpack, does a little setup necessary for mocha spawns the browser you want, injects everything and redirects the output to to the console. So basically, when one of the nice things here is that we get the same output that we get from karma from mocha, so we look at the same type of formats and formatting, highlighting everything is the same because it actually comes from the same place.

E

This is basically mocha running in the inner browser. Also, the source marks are a little bit better than the ones we get from karma you.

E

You can actually jump to the correct line. We get the nice stiffs. We had everything the same way we would have if we were to run mocha in an old apartment when other nice. Nice thing is the the debug mode she's a little bit Stabler, let's call it that then Karma's. So we can debugger thing. We can look at unhandled exceptions and caught errors and stuff like that, with dev tools and debug. Everything, and with this way the browser window is kept open, we even get the favicon from mocha.

E

Another thing is that you can run this stuff inside a web worker I, don't remember the.

C

E

Okay, so run with more water mode.

E

The tests are run inside.

E

So basically, you see here the worker running and everything is run inside of this worker. The same way, we have right now, good karma, so new stuff that we don't have right now is the ability to run in incognito mode.

E

So you get a new window, you can check here this incognito and everything runs the same way. You can click here and you get the file correct file can develop directly here.

E

So another thing that it's new and Wael loves it is. We can also run it in extension mode and the extension mode has a new feature which is actually really useful.

E

Like an hour ago. I made this. It automatically opens the dev tools for the background page. The background page is basically the environments that the extension runs, and this is useful because if, without this, it's kind of troublesome, you need to get to here, click here and then enable developer moles to get this nice link and to get finally to the stuff you want, which is a dev tools for the correct environments, and now it's automatic so basically I'm using the.

E

Playwright's and I'm simulating clicks here, clicks here and redirecting to this page, but for the end user everything is automatic and easy, so yeah, that's it one. Other thing that this can already do is run tape. Tests right now he just doesn't know when to hand and the process. So it keeps everything open. You need to first close it but I'm working on it, but it's already runs because tape doesn't need anything any setup you just. We just need to bundle everything.

E

So it's it's able to run with the mocha environment and I'm already doing the the the tape runner here without the mocha stuff. But even in the book environments, everything runs and you get the errors you get the source maps and the same output. You would get for running tape in another environment- yeah, that's it!

E

Hopefully, I can get this all wrapped up and cleaned up and use it start using it, and if we don't I, don't find any major issues, maybe incorporate it in Asia and finally, remove karma I still need to figure out how to not force everyone to download a million times playwrights, because it's big it downloads, three different, browsers and I'm, still researching a little bit. How can I make it easier on on everyone's laptops and hard disks or SSDs and stuff like that? But I know I'll figure a solution.

E

I'll figure out the solution, because right now we already have electron inside a year. If I had this, it's gonna be pretty huge.

E

Either we make every week every repo into a mono repo and we only need it once or I need to figure out the solution. Yeah. That's it hope. You like it I.

A

Like it, it's super cool cuz, both the balls incognito, even though it may sound weird. Why incremental matters more and more additional called surface in browser browser vendors lands for incognito mode, mostly around, like tracking, like websites, detecting like websites, detecting that you are running incognito or websites, detecting you're running like ad blocker stuff like that. So even though those are ball, even if you incognito is like regular Dom context, it may like behavior, maybe on the edge cases on some like fringe, IP eyes may be different and we tend to use those fringe.

A

Ip is in our stack, so it's pretty good to have it out of the box like part of the testers and for browser extension is pretty time-consuming. As you could see, it takes like five clicks just to get to the console and in chromium there. As far as you know, there were. There is no like shortcut in Firefox. You can do that. However, like okay ate over 80% of the staff users and complex companion, users I think that most of you Chrome, so it makes stuff much much easier and also the browser extension context.

A

It right now it's a regular web page like we've done on an additional API web extension. Api is on top of that, however behavior or around some things may be different and having that super easy way of writing tests and like checking if we don't hit any like hard-coded limitations of those contacts, will be super useful going forward. I think we talked early privately on that, but the overall plan is more or less for those changes, tool and energy right, either as a dependency or an option just for yeah.

E

I just need to figure out the the downloads part, because I don't want to increase a here. We play right, it's it's already too big and this one, even though it has some nice stuff in it, and it's so simple and stable, and it's so much better. But it's too big because we we basically install a here everywhere.

E

In every point, we have a bunch of repos in our computers, so yeah I need to figure out a way to tweeze that transition, but yeah and I just need to figure out that part, but you can still look. We can use it like installed globally and stuff like that. I'll probably ask some of you to test it out and see if it's any issues already tested in our repos and it runs it says normally so yeah we'll see.

A

E

Any and it's also as a watch mode that basically spawns new pair new tabs instead of just reloading, which is a little bit safer than then just reloading. The other page.

A

Does anyone have an additional agenda items because I can add one ad hoc.

A

I've been working on like redoing dogs, for the subdomain feature and just like a sneak peek take over all documentation of a config file, got changed recently and I made the Gateway section conform to that. It's not in the master yet cuz it will land with the subdomain feature. However, now it's much easier to like hot link to each each part of the Gateway section, and we got this like global toggle, for we will have when it is PR ones for disabling DNS link.

A

Look up look up globally based on the host header, so other people may point their ideas, their domain names at your key, so someone some people may want to disable that behavior and then this new path, prefix section, which now will have more substantial examples. So, for example, here like each parameter, has an example with some.

C

Thank you. Let's see breaking up laughs everybody else. Yes,.

D

C

You see we're broken up.

C

You're speaking, a 56k.

A

Yeah I realized I plugged my laptop and he probably started throttling low battery or something apologies for that. Was it like entire, like a segment choppy or no.

C

You picked off, maybe just just above this.

A

Alright, so quickly the Gateway yeah, the gateway configuration in this section has more examples with additional context. I try to add an example based on, like actual gateway that lives somewhere. Those are like our canonical gateways, one for subdomains or for paths. So it's much much easier for people to understand those configuration options in that context and then at the end there are some like receipts for specific types of gateway that people may want to run with both examples and additional context.

A

It's on the PR, which I will link in notes in meeting notes, but that's more or less where we are with that. Next step for the subdomain business is having selected matrix for proximal cuz, there's like a regular HTTP proxy mode and there's this like HTTP tunnel mode, and we support both. So people will.

A

It will just work, you never know which most people run and people honestly I did not fully understand the difference and until I had to like and some stuff around that so I like we will just internet everything. The the cave up here is like I. Don't want to grow the test too much because you need to like eat.

A

For example, if you want to test request a subdomain, we would have to run that one directly once like twice in each proxy mode and if there's a DNS link resolve in the picture somewhere that that's like additional permutation and I want to like collapse. The number of permutations that would probably write some other additional orchestration for those gateway tests, but so far it looks pretty promising.

A

The goal is to have it like the surface, to keep the surface small and easy to maintain, because it's like even for me, really sitting tests after a week, it's no fun and if someone will release the DOE test in few months, I want them to have like at least some fun like appreciating someone care but yeah, that's about it.

A

Any questions Roberto, oh man, I hope the nicknames are not stick.

A

All right, I, don't believe we have any more items this week. We won't be meeting on this call next week because we will be meeting hopefully face to face. So this call will happen in two weeks: I guess. Unless we want, we have any like open topics you want to discuss. We can just cut it short this week and return you like twenty minutes of your time.

A

Does it sound good? Yes,.

C

A

Alright see you next week and see us in two weeks, bye.

D

A