IPFS Virtual Meetups, 30 Jun 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: What's new in js-ipfs - Alex Potsides

Description

Don’t miss this update from Alex Potsides on all of the goodies js-ipfs had recently added, including Bitswap 1.2.0, much faster file transfers, and go-ipfs 0.5.x compatibility!

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

Okay, so this is a what's new in gensokyo tests: can it just a little overview of where at where you've been and where we're going? It's me very exciting, so this is me I'm Alex, platini's I am taking brain on most Internet's. You can find me on Twitter and github, and all that kind of stuff I work for protocol dubs and I am the lead maintainer on just like EFS. So what is Chasse make it fast?

A

In case you don't know it's an implementation of AI, professor and all the supporting protocols completely ruin in JavaScript, most Orion in JavaScript. There is some wasn't in there. Why not yeah I, guess about 5,000 own leads a week, and it runs in places most of the places that Jerris runs so in the browser, which is amazing, no electron and the typical focus has been on places that God gave us can't that's.

A

Why we're so keen on improving the browser, support, hello, loads of people, building really cool stuff on top of Jaso PFS, of course, notably Microsoft, who launched ion, which we've talked about a little bit and you'll, hear loads more from Daniel. Later on, and so what we've been doing, what would be happy? We see we've been trying to ship things.

A

So if you look at the kind of the releases that we do, we've done ten releases and the last twelve months, they've been quite a few point releases in between this bug fixes and that kind of thing.

A

But these are the ones that deliver new functionality, as you can see we're trying to trying to get the number of days between releases down so that we try to avoid big bang releases and do sort of a release, early release, often kind of approach and small digestible chunks, because nobody likes when you upgrade something and everything explodes then a few little things which play so.

A

What, if you actually shipped I mean we should layers of stuff I'm gonna go through all these lists, because there's too much but yeah, so you can see like going back through may I things UNICEF SP 1.5, if that was enormous, Maison carrée balls, that was enormous, is a really good, fun stuff, so a portable request. That was something that landed recently.

A

That means that you can now time out requests and I bought things so like if things go into your bits or one, this then Accenture and you cancel the request and they get removed because some of the big features the ace English for refactor. So we have this. We had this API that used pool streams and it was it worked, and it was great.

A

It was quite hard to use for beginners because pushes on a native feature of JavaScript, so you have to learn the API of the of a whole ecosystem of modules and I just made life quite complicated in terms of debugging, because the stack traces didn't really have all that stuff. There was very much there was very useful and everything it was basically a grand re right. So you know on the ape on the surface, it looks like just a few changes, but actually all the way down.

A

It's right to the networking stack to a very basis like their p2p and now, like the amount of code that we managed to delete by doing this is incredible. So it's much more than an easier to debug.

A

Now unix the recipe 1.5 is a feature that we shipped this adds metadata to unix FS files. So initially it's just the foil mode and the modification time, which means that you could.

A

You could mount this as like your MF s or something you could mount it under fuse and have the have a native place within ifs just for any file, modifications, executable and you know, readable, writable, no kind of thing users, familiar tools, just touch and chmod to make these changes to the files, keeping it small we're trying to we're trying to make it small. Like I, say we we run in the browser so we're very sensitive to the bundle size.

A

The barman size is quite big. So over time you can see we've.

A

Not really it's about the same size as it was. We had this this bump in the middle, which is a bit of, is actually just a Miss configuration, but we kind of went up and then we go down and we're always trying to to make it smaller, because it's too big it's way too big. Why is it so big? This is why I say big. If you look on the right this sauce, that's kind of that's actually jeaious ipfs, it's actually a tiny, tiny section of the bundle.

A

Most of it is things like note forge for doing which facing me parties for a little p2p for doing all of the encryption when you're sending stuff backwards and forwards on the network. I mean you see, there's there's a few implementations of it's a big number and bnj yeah, so they basically very similar things and there's also JSP n in node, forge so trying to target these kind of things, for a mute for removal and refactoring would be great, and so like javascript now has big numbers right. We can actually use big numbers.

A

Basically, if, um if you look at the documentation for this, the first thing it says pretty much is not suitable for use in cryptography. hahaha Why why why? So?

A

We still need some kind of way of doing like handling numbers that are bigger than the the size that JavaScript can address, which is super tedious peer, so we're forever trying to remove dependencies and make these things smaller, because because it's basically this the node modules before that- is that one of the heaviest objects in the known universe, well, sweet Anya, we're trying to make it fast, then quite a lot of work trying to improve the performance of jairus ipfs, because because jess is slow right, I mean, thankfully, most of the things that we do are actually network bound and so the the slow thing.

A

The thing that we can't we chose faster out, really is doing CPU, bumpers and but we can make other things faster. So he compared the transfer speeds between jsr, PFS, 38 and 46. The 46 was released in last week. 38 was a year ago. 36 was a year ago.

A

I, don't have standards that you are SATs of 38 and between 38 and 46 we've seen a 600 percent speed increase depending on the size of the file and basically the the bigger the phone gets, the bigger the income scale, which is incredible and we're pretty happy with that, and the reason there's a few. A few reasons why this has got faster.

A

One is we've adopted bits for 1.2 which allows you to prioritize sending blocks to other peers and among other improvements, but we've also spent time just optimizing the existing implementation, so we just identified some bottlenecks and remove them and we've seen an enormous speed up and this we're just getting started just getting started and other things we've been working on is the time it takes to import files.

A

So if I run this benchmark earlier today, adding a directory of two point: three four gigabytes of files between two hundred points and 200 megabytes: okay, sorry, invest 36 verses, 46, there was a 22% speed increase, which is quite quite impressive, and also also, if you look it's a little bit faster than go ipfs, which is fantastic. It's most of e. That's me faster, it's something so where we going next what's next, so next up, we've got a watch group support like this appearance.

A

The number one requested feature which there are so many types clip shops out there who really really want to use typescript and we're so we're adding air the edge of the library so we're not because of like the kind of the tourism that is, that every every type of developer news JavaScript, but not a JavaScript developer, knows typescript. So we don't necessarily want to convert the whole library to tighter grip because, like it lowers, it makes this the pool of potential contributors smaller raises the pirate of entry.

A

So what we're doing is we're going to typescript definitions of all external interfaces, so the people coding against oSIP FS can use it as if it were a native typescript project and we're hoping repeats getting that out. What's going to come up this quarter, but I think it slipped a little bit, but it's definitely very high on the priority list.

A

Also react native support. We really want to get chase. Our vests running react native, so they can it's a it's a very commonly requested feature we want to be able to like. This is really interesting. Ones are sharing a running, though, between single origin browser tabs. If you have a website and it's using json AFS and you open a new tab at the moment, it has to spin up a new JSON if first load, which will have you know possibly has the same parity.

A

And you know you have to make it there's network connections again and it's using the same storage, and this is kind of bad. So what we want to do is move and I could first node into a shared worker or just allow the website to deploy in a shared worker, and then you will be able to access the shared worker from all the different types which will reduce the amount of like CPU using a network resource, and that kind of thing then also they could not actually be on the show worker.

A

It could be actually using a proposed desktop. And if you, if you established that kind of communication channel between your web app running in one or more terms and a central node on your machine, then it makes actually then adopting those changes. Much easier. We're gonna refer to the pinning system because it's it works, but it's quite slow and the performance gets worse over time.

A

So we've got some plans to move all the pins into the data store and our early benchmarking has shown some very promising numbers which will bring that file import time down even further, which would be amazingly good, DHT we'd love to have a DHT. At the moment.

A

The gorilla this team are massively refactoring, their DHT and at some point, liver will stabilize and there will be a specification written and then we will implement that specification in JavaScript and they're kind of, like the last, the last missing bit of functionality really to be a fully fledged participant on the oculus network. I guess it's never really been a proper priority before, because because I've focused on running in the browser, most browsers nodes are not diable.

A

So there's no point saying to the DA: there's no point: advertising the providing of C IDs on the HT. If no one can connect you, so it's never gonna been a high priority, but more and more people are using it in node at that point becomes very important to have an implementation that is stable and free featured onwards. We have a road map. So if you follow the euro, it's one that I'm going to read it out because bf.

A

If you visit the road map, you can see what's been slated to appear in upcoming releases and you can see like where you can help which this leads me on to say. Thank you to all our contributors. So here's a list of all the people who have contributed to there's just an orifice. The J's I'll give it me PO. There are plenty more across the ecosystem and it's amazing. So thank you all very much. You can make that project. Isn't it?

A

B

B

Awesome, thank you very much for the update Alex just quickly. Looking through here, we've got a couple questions for JSF BFS. What do you see is the relationship between JSI, PFS and rust, IBM, as given that rust can produce web assembly and run in a browser? Well,.

A

I mean is great that there's more implementations of like FS out there I think you know I think each implementation is, can I do things that make sense to the platform that it targets. So you may see like an idiomatic API use from Ross. Oh yes, where is the Jess? One will always try to be the closest to you know what you would expect as a JavaScript developer, but I mean it's great great.

B

Hassle and next question for everyone who's using jess ipfs at the moment. What's the best recommended complete typescript effort to use right now until official TS support is launched so.

A

I think definitely type has a bunch of definitions for the interfaces already I mean I. Think that's one, that's what you were doing.

B

Great and have you given some thought about doing remote requests like proxy requests to be a running desktop ipfs? Yes,.

A

Yeah I mean I mentioned it in the toaster. If we define these kind of interfaces between the running code and like the API, that's expected on the remote, the remote node becomes much easier to then adopt those kind of things yeah. It's definitely something that we're interested in looking at so propose things on the repo great.

B

All right, I think that's it for questions. There's one from Stefan, but I'm gonna leave that one for a Dean in the Go ipfs talk I'm talking about new metrics, but thank you very much Alex for coming and talk to us and if you have any more questions for Alex feel free to put them in the chat or in the QA, and we can answer those ad hoc.