IPFS Weekly Call, 27 Nov 2017

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: IPFS All Hands 🙌🏽📞 Nov 27, 2017

Description

See https://github.com/ipfs/pm/issues/528 for details

A

If you're recording, then okay, sorry I'm Beverly.

B

Hills no wait.

A

Okay, this is the All Hands call for November, 27, 2017 and first on the agenda. It looks like we have benchmark J s ipfs from David.

C

C

Okay, I just see my video really interesting. It's.

A

A little bit choppy yeah.

C

It should not be here all right, sorry, no more distractions, so benchmark. So one of our computers made a really nice website that basically lets users check out speed up my PMS. If you go to benchmark Jess that IP festered are you it got published today, so you are seeing the latest version, but still not the latest version with the next just appearance which is faster. You will find a website and me I can just like walk you through with a screen sure makes it easier to committed. Where is Mike?

C

Oh yes, actually, Firefox and abuse chrome anymore, so yeah.

C

Well, I was waiting for Firefox to catch up so yeah you get a really nice web page right now. It has like three benchmarks. You can start and he tells you it's using these you're 26 and then, once a note starts you'll see some peers connecting my peers connected and now you can run like one of these benchmarks. They have different files right now. It's like static files that yeah it's a face. It gives you a place.

A

C

Someone that I don't see a face. Okay, oh he's here.

D

C

You and I do to walk us through to the speech my platform.

D

Go ahead, it's uh I heard you talking there you're doing very well. Thank you.

C

Yeah, okay, so I was just like wrap it up and then you add more of your ideas and what you want to do this. So as I was saying you guys like really these guys like way of telling like how many blocks were there on the file like how much time did it take to get each work?

C

And you get like this read speed. So in a sense like this is kind of like a platform for more people to add more benchmarks of streaming files or just like different kind of network situations and then compare it. We like real, just get HTTP such or even see it through time as it evolves and I've seen just at the offense cat master I was undecisive if the domain should be like ecfs fast, yet or benchmark just that I've episode I/o. But we went with this one because the sacrifical is already there.

C

It was simpler to set up yeah. We might get Astraeus two points, this one alone and so yeah. This period is much for the contribution by the way.

D

C

You so much for the contribution.

D

Thank you guys for inviting and I'm part of IP face. I can contribute in ipfs anyway, I hope, yeah, I, hope the benchmarks would be will become more useful from from you guys that you know ipfs much better and you.

D

It was this example was just get a stream benchmarking performance.

A

D

But I think we can improve it to become really useful. You know set of tools for keeping track of what's going on in terms.

A

D

Speed and what's going on the background.

C

Yeah so I guess, like you, are jumping into the next point in the agenda, if I recall correctly, which is like more benchmarks, so we create the issue here, just like again, I guess it was created today we kind of like publish this today, although this was not then like some weeks ago, the the core of the work, and so the issue is about like getting feedback on like what will people want to see on this website? Is it like having more files to be added?

C

Is there like any kind of visualization that would help everyone to understand what is going on I, wonder every minds just getting like a big square like I. Think transmission does this, which, like you, have a big square with all the tiny box and I guess as a blogs are coming? Are they are being pipes to your nodes? You see the blocks appearing on this big square and finally, your the whole piece. So we can do more visualizations, we can add more tests. We can compare it.

C

We can benchmark it against HTTP and and see the differences right because, like when it's like a small file, poor HTTP is going to be faster because that small file comes from single endpoint. You don't have to do any crypto magic just like downloaded, but then we got like a large file like a two gig file. I thief. S is certainly going to be faster and and having those comparisons going to be useful and that's it I think I can stop my screen shirt here, Hey any questions, ideas.

C

Anything stay here, Jorge also something.

D

Else well well, yeah, and we can I, don't know much about you know. What's going going with ipfs connections with fears, you know there is picture that you mentioned. There was really good to keep track of what chunks are you getting from home? It's it's a really good feature, I think also, and also we can. You know we can start with some basic stuff that basic matrix, that for those who use ipfs, would be really useful and we can start writing them down and the implementing and yeah I think in the process.

D

Many things will come up and yeah make it more useful. I wasn't expecting to be such a so much useful in beginning I just created it just to show the speech, but oh.

D

Yeah I'm in the discussion in the issue I mentioned also Victor's, some of the some of Victor's comments on some ideas. He had about running multiple tests for each get.

D

A

D

To keep the medians have a much better understanding with multiple tests and the functionality of pinning and being files, a which I think it's a good idea where you can also have you know a list of your files locally, and then you know, try to play around and see how the speed goes and also um yeah.

D

How many sorry for my english, guys, I, think I'm, torturing you but.

D

We can also have a better understanding of this browser class, which say it's a mystery, because you know III was trying to debug it and trying to you know see what's going on, but I could never. I was never sure of. What's going on in the background, so yeah I don't know, maybe if we run tests multiple times this. This issue over this bug might come up more often and.

A

We better understand I, don't.

D

Know, I haven't a single process. I haven't seen the relative issues if it's totally solved, but I have experience recently. Also some browser cache misses, so we can, we can get. We can investigate more detail, I, guess and also.

E

D

We can, we can make it like, look like more it dashboard rate and then add more and more. You know benchmarks for getting files benchmarks for boots, for adding files and.

A

D

Kind of approach, any ideas are welcome and also from you.

D

You know creating a better benchmark from it could be really useful only from those who really use it. You know and I want to have the numbers that have a meaning make sense, and so that way, if we write down the specifications and it's it's more targeted and I- think it would be more useful.

D

Okay, I don't want to hijack the conversation.

C

It's good yeah. Let's continue on the issue, there's a lot of ideas and like things that we can, you can use a benchmark so different problems. We can use the benchmark just like to see improvements in performance and memory usage and yes, I really agree with the dashboard, because if we have a dashboard, it makes those tests very repeatable and very very fast and like it, empowers everyone to actually participate on that debugging and figuring things out with this.

D

A

Alright, if we're ready to go on to the next agenda item, it looks like we have live period, appear transport from David.

C

All right wait: wait if you want to mix up like a chef, all denied wins just go for it: okay, I might I might just like click enter and so like it's like, like there's more people, so, okay, I'll just do this one new. If you transfer a WebSocket start and I thought it would be good to mention that now it's ready for program, you can use it. In fact, the next version of Jessica TFS will have it installed by default, and then you can just activate it with the multi other.

C

It should be as simple as I'm doing any, enabling any other transport this web socket star is very similar to web RTC star. So you got a rend of a point where you died out you and then you use this friend of a point for discovery. So like any know, that dial set point will also get to know that mm at some point.

C

So it's kind of like a multicast DNS of the browser and the difference between web RTC, star and web socket star is that web socket star uses a web socket end point to relay the messages while web RTC just chooses that end points to relay the signaling data that then what you establish a peer-to-peer channel.

C

There are a couple of reasons why this transport was built like the major one is really because what you see is very memory hungry and it's really hard to control, and there are things that we know that we can do. But this was kind of like a short thing, like a quick thing that we could build in order for people to have the same experience that they have with a weber ec star without having to wait more weeks or even a couple months to see a full round solution.

C

But now that control the beasts, that WebRTC is in essence, WebSocket star. It's kind of like circus relay. It's very similar, like it's almost like a different color of the same thing or the same functionality, but implemented a different way, because circuit rely enables you to use any ipfs, no to relay your connection to another appear and that you have to have the multi other of that peer as a secret, real-life multi other we, the WebSocket star, you are still relaying the connection, but it is a transport so like in the design of a peer-to-peer.

C

The design was purposefully made to support that I get transferred or any other piece of the. This stack can be like a set of modules itself and so realize that transport Ned knows how to use swarm and stream muxing and second you to create an order. Transport in case of WebSocket Star is something more low-level, more simple that just uses a WebSocket in points to route connections, I, guess in practical terms where the screen is like it's a transport that doesn't do the crashing thing that people have seen.

C

We have detected some bugs on it today, like it's, not very reliable, on closing peer connections, so you can dial you can discover, but if you go off and on where a lot of times the render who point gets a little bit confused, so it needs more tests there and and yeah and I'll see you because I saw a conversation on the IRC channel about the fusion like this is the route that you want to go. We want to create these wonderful points for our transfers, and the answer is no.

C

This is like a short-term solution. What we are going to do next is react to texture the way that module loading works on the peer-to-peer so that we can have by this common. It's like common vertical in some repute appeared that enables any piece to use all the other pieces simply right. Now, it's not trivial trivial because of the way that dependencies are required and injective, but like once that really texture is there.

C

Then a transport will be very able to call something from upper layer and- and so that means that I can set up in one single render point, then any node in the network can become a renewable point for any other nodes to meet again very much like circuit, really just different colors of the same thing of the same product and then we'll remove those central points which, yes, we don't want them. We don't need them. It just like a question on really texturing things.

C

So that's it works slightly and then, if I still respect to them. So on.

C

Yeah I hope this explanation was useful, maybe not everyone at the full context of all the things you know feel free to ask me questions now or later, or ask me for pointers.

E

Yeah to be, could you maybe after we've discussed, can you make sure that update them notes to have things to like what is WebSocket star and what is what RTC star I'm good like? Where can people read more about this and what and also like, related or requests or issues in the pupae space.

A

Alright, we're ready to proceed on to the next agenda item um pick one. It looks like ipfs parcel only take two from Victor yeah.

B

This is just a quick note: I put the link in the in the notes as well right after its plan to be the the 16 of the summer. He will be kind of the same format. Hopefully the high time will be a bit better, but if you are close point Barcelona and you want to come- that will be very fun or if you want to participate or help out remotely. That will be super nice as well.

A

All right, it looks like also above that, one uh this week, JSI PFS 2.7 once released and there's an issue for that. There all right.

C

Just a bit, so the release is ready. Finally, we did everything if we wanted to do and the notes on what is coming. We have to really czar on that issue and now that we discover that issues with WebSockets star we're going to fix them, but the release should be published in today.

C

So Wednesday, if you are a user of just a TFS, please do tests this latest version and let us know if it works for you if you find any problem with your application with your tests, and it is always like very relative to see other people run their test Suites against the latest, because just gives us like so many like. Sometimes it's just a level of like a degree of confidence that we know. Okay, like things are working and it's great all the times.

C

It's actually they find bugs, and that's also great, because then we can fix them before breaking people. Software and yeah.

A

Okay, um off of that, it looks like David, you don't mind. Continuing the the web, UI and station are getting lots of love, I'll.

C

Do it so this is kinda the edge and he's a new contributor to the whole project. He started like weeks ago, just like going through issues and and submitting for requests, and over this time he managed to get web UI to work again.

C

I took tests on CI and to actually be buildable, which is something that didn't happen for quite some time like we need to update the web UI and go ipfs because, like the HTTP api of ipfs has been evolving just a TFS api has been evolving, but the web UI always never really updated. So it's kind of like behind.

C

If you are using web UI today, you probably have some problems so now the web UI it's already fixed, updated, it builds, we can get a new ash or that app and like updates the that is article dodongo, ipfs, and so thats users can use the web. um The other thing that's in the vacay also contributed and aki is probably going to join us next week with the demo, I'm just like, sharing it bytes, so yeah see station, so station works again and it's all updated.

C

So you can now, and I will also have links to both of these things. Don't worry if you are not finding it now, but station is in electron app and these Styles, like a menu bar item and it's a nightly fest know that you can drag and drop files and, like then, it copies to your clipboard, the ipfs at io / / of that file.

C

So you can just share with friends anyway, light turned on and off your note, it's kind of like a nice way to install an IP, fest node without having to do terminal command line things and in second eye security, and there are some big ideas of like how that thing should look like and how it should work and which features you can have.

C

So if you also have ideas, go to those repos and- and he shows with ideas and and hopefully in the next couple of weeks, / next quarter to actually get something that we were. The idea is to walk to go towards a scenario where we have them. The name might change, but like the ipfs dashboard, which, like a full page like the web UI, but like with more features because ipfs has evolved a lot since for you iOS design, and then you have the ipfs UI for like menu bars or for Firefox and Chrome extensions.

C

So like when you install the ipfs companion, you get like a tiny UI box. When you saw the electron happy, you get a tiny eye box and like if we can module eyes those UI pages web components, then we can reuse them across different integrations to the electron, to the desktop to the browser, one with EFS companion, to a view I, which is a larger, that's world, and also because there will be web components. We can enable and disable them, depending on the feature that the user wants to use the features.

C

That's the IP personalization as that they are using with, and also enabling the user to say, I want to use an m-pam, which is a defense or I want to use a remote node with my PFS. So yeah there's a lot of like really good ideas in terms of like getting users into ipfs and I start playing with it.

C

A lot of it will be around these repos like FIFA's, companion, web UI and station, so stay tuned, and also, if you are interested in reading, just write a note in the notes of this session or in the ripples.

A

Alright, it looks like we have the critical issues for go: ipfs and JSI PFS, so.

E

I threw that in there just just a surface that we have these lists, that, during this quarter, sort of behind the scenes, bigger, ipfs team and JSI PFS team have been putting a lot of work into addressing known, critical issues and there's nobody well Hector's the closest thing to the NGO ipfs team today in Dimitri, I guess but um Jeremy's on a plane and I think Kubas in a class right now, but we could at least look at the NGO ipfs ones and just that some of them have been addressed already so connection closing, which was a long-standing challenge, also swarm hot pants, optimized I'm, not sure what that one is.

E

We don't have Steve in here and fixing some flood sub stuff, so you can check in with that list. I would say next well, I won't be here next week, but maybe on next week's call. Some of the NGO committers can look through this listing, give an update of how the progress is going on those ones and then, unless anything jumps out at anyone about the good stuff, we could just look at the GS ones.

E

Okay to me, do you want to give a quick glance at the J's ones? You've been hammering through yeah, absolutely so I'll just share my screen again.

C

All right so I used to alpha board to track all of the shoes that we prioritized for this quarter and by the way, I, don't know what happened to zoom, but the window with everyone's face just disappeared, so I am with really no feedback. If someone raise your hand, please tell me: okay, we have here the waffle board, it's already filtered by piece, you're critical, and we can see that there has been a lot of progress. We started with 51 issues on the ready column and right now we are 17 left and five in progress.

C

The like a lot of performance issues have been sold in Jessica fest and like bugs that were just annoying that were there. The next two really big items that still need to get some attention are the IPL de track. So there was a lot of development or a lot of improvements proposed by the end of last quarter and like there has been killed.

C

There was code created to fill the needs that some users were reporting, but we never got the chance to actually just push those changes through the entire the entirety of the IPL D stack. So that's the work here. You see support my API and an IPL de resolvers like this is a bug and then the IPL, the resolver changes also attached to this one. One of the issues it was critical for GSI PFS is getting the back API supportive on Jessup, EFS API, so the client that typically people use with go ipfs.

C

This work depends a lot of work that needs to happen on go ipfs, land and yeah. We essentially need to revisit this like see if the design that we decided on and like the way that we decided to approach this are still valid and if there is something that we can still do this quarter, so the other big piece is really the DHT work, so we in Jessica's land, we have a DHT implementation, but it's not interoperable with the girl. Yet, although it works in jest lands and yeah liking, it needs more debugging.

C

A lot of the pieces are there, but there is something some message that is not passing and it seems more like a. We pretty pure problem, then necessarily, and she has a PFS problem, but it detects al Jessup if s works and the experience that users get so for the remaining of the quarter again just to to emphasize it will be things around by PL, D and then DHT by that order.

C

C

Yeah, if there are any questions or like what this is mean, I'm sure like I, will get these issues and, like I, know what is written on the whole trip, so I'm seeing it but like if you have a question on this status of something you can trust now and yeah, go ahead. Funny.

F

So talking to Steven right I think there is this possibility of resolving IP LD using the Gateway. Is that still a possibility, or is it going to be really through of the API Daggett? You.

C

Mean by having a / IP OD / CID into the like just doing ipfs yeah. We definitely talked a lot about that, or at least in multiple occasions and like no one is against it. It is.

C

It is more a question of would that be useful today, like what people actually need at API and so far the answer has been no, and maybe you are going to tell us that we are wrong because you have some, perhaps I use case for it. The reality is like most people that used a gateway one to access files. They are not interested in like single nodes of graph and like when people wants to operate over graphs.

C

They really want an API like something they can create a program and a critic graph and perverse with and I do filtering and selection and transformations, and so on. So so, if there is a case to go into chair, go for it well,.

F

Not really so, actually my IP id specification, the D ID spec on top of IP FS I, built it using IP, LD now and so I think so I'm, mostly using the API, which is a little bit in flux right now. So I think it doesn't really matter I'm building it on top of the API, but I think. Ultimately, it's going to be for other people to resolve for, like the sovereign foundation, to be able to just hit the game gateway to resolve the the object and so I think mostly for convenience.

F

For other users, not myself, you see.

C

F

Guess the entry really is about resolving. You can get the JSON document, but the JSON document has to be serialized in Seaborg and you have to validate the hash of the content. I guess is really the issue mm-hmm.

C

F

Sure it's deterministic, it.

C

Is really hard to give like a standalone, HTTP interface, so ike without any client in the front for IPL be so we can totally like show you data.

C

We can patch a note and I give you a JSON version of the data, but as soon as it has like data types that are offers or like different kinds of numbers, because the way that Jason gets string defied and put in the wire it's not like messes with that data, which means that, like you, might be grabbing some piece of Jason from like / IPL d, /c ad and any relative jason, and then you say lies it again and like do this IV and it's not going to match the same city if you were fetching initially, so this is kind of like a really hard problem.

C

The the solution for this would be just like creating maybe like, instead of you getting their raw nodes, getting you a visualizer so like when people load something on / IPL d, they actually weld something I got a playground.

C

If you have visited like that seaboard or il playground or Jason lb playground, you like this box, it tells you the data that's inside and this analyze the version so that, if you grab a stylized version, then you can reproduce the same content locally, using pipe Affairs or just a few pairs or in the future fire papers and so on, but yeah like I guess. We just need to understand what are the needs and like what is the best thing that we can put on that path to make it useful for users.

F

Okay, cool yeah I think we ought to start spinner as an issue. I think not high priority.

C

Yeah but like if there is any endpoint that you'd need in order to get the IDS moving forward, just also yeah, what I figured it up.

F

Yeah so mostly I've been able to do it right through I was rectum. Writing it into ipfs the library but actually realized that just using the API is just fine I guess there was the interface go, interface, go I, a P, I and so I'm, basically building on top of that and so key store and everything else works. Just fine, so I'm building it basically and it's the D ID spec is is basically modeling of JSON. It's a pretty straightforward.

C

As a food, okay,.

A

Alright, it looks like next: we have a pull request on JSI PFS decontrol, requesting some feedback.

G

Just where there has been some talks about having a consistent module to start and stop ipfs notes and that's basically a proposal for that and trying to give some feedback, because that is a blocker for getting the required amount of testing on on super really to get it ready for primetime. So that's kind of an important piece of that as well so yeah.

A

G

Feedback on on the proposed interface, if there's something that you know, requires comments or criticism, leave it there. But I am at this point kind of trying to time box it for the next 24 hours and you know: there's no protection step forward, but I could both. But if there are, let's, let's make sure we'll bring him up, yeah, that's prevention, all right right! Every time, yeah.

C

So um so me and Mitch we have been talking about- is essentially just to give everyone a little bit more context is ipfs D CTL.

C

We call it the demon controller and it is the module that typically people use from json to spawn, go ipfs demons and because now there is the need to also spawn J hyperfest demons. To do all these job testing, we kind of like figure out we better to just create a module that it is more flexible, more versatile than creating a different module and the demon controller module.

C

It's kind of like all the in terms of interface like it doesn't support a lot of things, and it's also like not very well documented and and so because we learn some of the UX that developers expect by building what we call the ipfs Factory. It's the thing that we have for testing on Jessica, the Sofia in Jessica verse. We figured I'd like okay, let's just like bring all that innovation back this module and make it available for the entire community.

C

Like then, we have a single place to spawn nodes from electron, from the browser from Mike, single javascript file, etc, and dozen demons can be just rivers can be a girl or in the future, any other implementation.

C

So the question in the house here is: if you use it to spawn demons for your code and, like I, know other applications. You should definitely make sure to check that full request, because it has a new proposal for its API, which should be simpler.

C

It should be nicer, should be more debuggable and if there is something that, like you, created on top of the old API that you see that we are missing and you think will be better to also add- definitely bring it up or if you see that we are removing something which should not be the case. But if we see that we are removing something that will break your usage of it. Also, let us know- and that's that's convention. Otherwise, like again, we have the PR open.

C

It will get a little bit like some days to get developed, so you still have some time and if we have to add something in the future, we can always add more. It's fine.

G

Yep, absolutely thanks. Eddie, no problem. Thank you for pushing any.

A

All right, if we were ready to move on then looks like we have a demo on the current state of Jenkins from Victor. Yes,.

B

Hello, so everyone can can hear me good, yeah, perfect, okay, I am gonna share my browser, which happens to be Firefox. It's always been Firefox I'm, not a traitor. So let's see we have this long-running. Now everything is right here, but don't worry it's my favorite. So this is how Jenkins looks right now the is called goo ocean and he has like a new UI and things. So what I?

B

My my biggest focus, this quarter I have been working on Jenkins and trying to make our CI as good as possible, also involves making sure Travis and sonken CI and a player works well as well. So the the current state is that we now have.

B

We now have a repository, but I am under here, where we keep shared shared pipelines for all JavaScript jobs so far, so what it does. Basically, we have a couple of node.js versions. So, what's wrong, we have the OSS. We want to test some other Global's for the don't version, and then we have specific tests for Windows. As we know, it's a bit special and then we have the sets poor or UNIX compatible machine and I. Basically, we run through all of those in parallel, so how this would look no actual job.

B

It would look something like this, so this is speedy stream of Maksimir. For example, we can see that it took 35 minutes last four numbers which is kinda like too long time, so I'm gonna go over to one of the gory paper skills and so on. We can see the : pipeline. This looks like this. It's a bit it's worse because I just hacked it together. Yesterday, basically, but the same principle.

B

We have a bunch of Oso going on in Russia and we run through installing DX here, so checking out the code installing dependencies, rewriting your done testing and then how this would look. This is, for example, core IP for others. It's a lot faster. It takes like two minutes, so we can. Hopefully the demo gods are on my side. You can replay those and we should see it testing on all the OSS in the same time, so we don't have to wait. So the Linux version completed in 30 seconds, Rula.

B

One sex question: it's slower, and so is the Windows version as well, so how this would actually affect you on github. We can look up any time, so we have Co ipfs address, which is the first one. Oi I started with I didn't create a pull request here, but we have a branch, and here we can see the the status typical github status jumbo right now. This library only has one once here. Many projects have had multiple ones, so a better example would be.

B

Maybe one of the multi formats one so the the current state is all the JavaScript projects in multiplayer and IP on the year. I think has been done already and it's using using Jenkins.

B

So for our JavaScript project, you have the up layer, the circle and Travis file, and then we also put the Jenkins file in the CI folder. And then you just call this this function and your works.

B

So let's see playing the pool request online nice.

B

So now, every time making progress you get four different services. I know. I cannot see the old tons, oh yeah, so here we have circus er, admire Jenkins and Travis for only JavaScript project and I'm poor goal, I think it's Travis, Circle and now I'm, adding Jenkins, that's one. So here is a go flood sub, which I just added now as well, just using Travis and thank you and yeah- that's pretty much it.

B

The workers are one on Amazon and in private Mac OS cloud as well, so the AWS workers we can spin up whenever we want when we need more and we need more workers and then the Mac workers require a bit more work spin up. So currently we have one one Linux worker. We have two windows and two michaelis because.

A

B

Worker is so much faster than the other one, so we only need one this point and yeah so now. The next step is to just add this taken file in basically every single record that we have and then, after that, we can hopefully turn off the other NCOA services and everything will run much faster and that's why that's my update and we'll have any any questions.

E

So you'll be done like tomorrow, because there's not that many well.

B

Actually, it would just take like 10 minutes or something like, though I think I only have 300 records left, but I actually wrote a tool, at least for the Jas project, to sync, the files automatically I still have to manually check. Everything is passing and I found some minor bugs in the pipeline as well. It's pink on the go projects. I am not sure who.

E

C

This is really cool like I've wasted, so many hours waiting for colors, because you can always like try to do something else, but then you have also have to check in with province. If it has finished or not. Yeah, then.

B

You have to retry and.

E

I forgot now we're planning to turn off Travis in circle. Once we've got Jenkins running so those wanting it run. Yep.

B

That's the plan and that we can the planets also to add more architectures like arm to add different workers that are faster or slower. We can have different network conditions as well. We have we have a lot more possibilities of setting up, such as to run in the exact environment that we want to, rather than some hosted thing where they decide how we should run orders.

C

One quick question: when, or is it part of the plan for this quarter to get the GSI BFS API tests running against white papers all the time so that when there is a change in grab a plastic I've, even steam knows the GFI papist API I thought correct yeah.

C

So there was at one time that there was like a Mac OS machine running we teamcity yeah, who is using yes, I diversity, I tests to test with the go ipfs like the master, but that that went away, and so since that went away, we already found a bunch of problems like type offense. Just changed a couple of things, and then we had to catch up. Yeah and yeah. Is that coming back soon, yeah sure I.

B

Focused firstly, on writing a generic python for the standard, the standard jobs, the one story similar that can have the most impact in the shortest amount of time. But now I can focus on having the custom pipelines as well, which go ipfs also needs for itself, probably needs a custom Python not on generic on, and then it will be easier to add the JSI PFS api dependency. Well, the other way around the co ipfs build as a dependency for this idea. First api and then one the first. So yes, short and sweet.

A

All right: well then, that looks like we're at the end of the agenda. um Yeah and this concludes the All Hands meeting for November 27 2017.