CRDT and browser testing (Automerge & Polypane)

with David Mytton & Jean Yang

S01 E03


Automerge (conflict-free JSON-like data structure) & Polypane (browser testing tool), a devtools discussion with David Mytton and Jean Yang.

Episode notes

Episode 4 of the Console Devtools Podcast, a devtools discussion with David Mytton (Co-founder, Console) and Jean Yang (CEO, Akita Software).

Tools discussed:

  1. Automerge - JSON-like data structure for concurrent writes.
  2. Polypane - Browser testing tool

Other things mentioned

David: Welcome to the Console Devtools Podcast. A show all about interesting developer tools. I'm David Mytton, co-founder of Console.

Jean: And I'm Jean Yang, CEO of Akita Software, an API observability startup. In each episode, we'll discuss two interesting developer tools. We're keeping this to 15 minutes, so let's get started.

David: Our first tool this week is Automerge. This is a JSON light data structure for concurrent rates. It's an open source project and essentially allows you to build against the idea of local first software. This was an academic paper that was published from Cambridge University a couple of years ago, but it essentially builds out the idea of a conflict-free replicated data type or CRDT, which means it makes it easier for you to build in versioned access into data structures so the data can live locally and it maintains a history of all the changes, making it very easy to then sync. And the goal for this is to allow collaborative text editing. It allows for the idea of functional reactive programming. The reason why I like it is the philosophy behind local first software, which includes seven different principles, which I won't go through all of them.

My favorite of them is the ability to work locally so that you are storing and controlling your own data, and that it doesn't have to work with a network. And the reason for this is the ability to control the data that you're generating. A good example for me, personally, is my notes. I keep a lot of notes in plain, markdown format, and I sync them using the Syncthing. It means I can edit them, and use vim on any server or on Mac as well, but on Mac I can also use more native applications. It gives me, as the user, choice, which I think is something that developers should try and build into the applications.

Interestingly for this data type is that it's implemented in JavaScript, and so my question for you, Jean, as our resident programming expert, is why has JavaScript become so popular? Is JavaScript the right language to implement an in-memory data structure in particular? When you actually tell AutoMerge to enforce all the guarantees it provides, it has a significant performance impact. What did you think?

Jean: Yeah, these are great questions. I often ask myself the first one a lot. How did JavaScript become such a popular language? I remember the mid 2000s, I was just starting out as a developer and I was seeing the rising popularity of JavaScript. But at the time, there was still a chance that JavaScript wouldn't become the dominant language. I remember I had this internship at Google in 2007 and we would sit around at lunch and talk about what are the alternatives? What we really imagined was something WebAssembly like, so that still has a chance. But we were like, "Well, browsers just have to run something else and then it's okay." But it just, it seems like browsers run JavaScript, everyone writes things in the browser. And so JavaScript has become the assembly of the newest operating system that's the hottest. That said, it seems like many apps are becoming increasingly native to the browser.

A JavaScript based language does seem to be the language that makes sense for something like this. I guess the alternatives would be TypeScript, as we talked about a little bit before this, which fewer people use. I feel like JavaScript is just for reach. You could also do something like this on top of WebAssembly, but again, you're limiting your reach. I'm not familiar enough with the performance implications of using something like TypeScript instead . I’m not sure if that's much better, but I would doubt it.

The choice of language makes sense, but David, I had a question for you because the examples that you use make a lot of sense. It's great to be able to have local first Google Docs or local first something else, but how many developers are going around starting with local first as something they would do? How is this GitHub library for local first development so popular? Is it curiosity? Because this has over 10K stars? Are people just very excited about this? Are lots of developers using it? What are they using it for? Is it a commercial product? I was very curious about these questions?

David: It's very much an aspirational philosophy. I think we've all come from the era of files on your desktop, files locally. That has now moved to, not even files, with a Google Doc, there is no file. Even if you sync them using a local Google Drive client, if you open up the file, that's just a JSON snippet that has the URL to Google's interface. You get the advantages of that because it means everyone can edit on the very latest version of the document.

Despite Microsoft's efforts, Google Docs still is better for collaboration, for real-time collaboration. Just having those features that are available in the browser, it makes all the sense to start there, but then you do lose some of the control. You can't back up the files just because there aren't any files, they don't exist. You can't store them somewhere else. You can't archive them. As services change over time or get shut down, then you risk losing your data. That local first approach is really about giving the control back to the user,

David: But is it too much of an aspiration?

Jean: Yeah, that was my question. I couldn't tell how realistic this implementation was because they were using SQLite. It was one of those things where someone makes a cool new language, everyone stars it, everyone plays with it on the weekend, but people generally aren't using it in production. Is this particular library something that's getting used in production? Also, how often is local first getting used in production and are they using this kind of library?

David: It's basically an assumption now that you have internet access really, isn't it?

Jean: Right.

David: The files are available through whatever web interface you're using. There are very few applications these days that actually work on local files apart from Microsoft office, and maybe design files. But even with Figma, it’s online and you can do all the collaboration in the browser. The idea of having local software doesn't seem to be meet with the requirements of startups and people who are actually building projects.

Jean: It also seems like local support is something you do last. So, I'm curious about local first because I feel like if I were building a product that was a collaborative tool, I would build the in-browser, fully online version first because I have the most control. Then the more you do locally, the less visibility you have as a tool developer, the less feedback you're getting. It seems like it's sort of the on-prem equivalent of consumer software. Local first seems to be the opposite of how I would want to develop and test a tool.

David: I would say the developers will be used to Git, which is local first.

Jean: Yeah, that's the point.

David: You clone the repository, you do it all locally, and you push things, and that is kind of how local first philosophy is trying to approach things or at least with this library is providing an easy way for developers to implement that kind of workflow. Using  git as an example I think is probably a bad idea because it's such an unusual piece of software. The use cases are very specific to text files and generally development.

Jean: Right, I guess something like DropBox is also local first.

David: Alright, our second tool is Polypane, a browser testing tool. Now Jean, you've already thought about suggesting this to your team to adopt it. So why don't you explain what you liked about it?

Jean: Polypane is a really neat tool. The idea is that if you're doing front-end development, testing across different browsers, different browser sizes, different form factors, is quite tricky.

This is something I've run into a bunch with my company, Akita. What Polypane does is it's this new browser, that lets you test different panes side by side. You download Polypane and you get to manage different browser views all side by side. You can see how your page is going to look across different experiences all at once. This is built on Chrome, so the dev tools are available. They also have a bunch of really neat plugins, I admit I haven't tried yet, but they have accessibility testing, they have some chaos testing, they have a bunch of front-end testing built in.

We actually have no dedicated front-end people in our team right now. Quick plug, we are trying to hire one. So talk to us if this is something you like. Over the last couple years, I've done a bunch of testing for both our main site and our web console for our application. This multiple browser thing is a pain because there's stuff that looks good in a small window but not good in a big window. Sometimes for some of the things, they'll implement it assuming that developer is going to have a vertical window open, but around my demos, it's actually a horizontal window and it looks completely different. So, we do work a little bit with a designer who helps us with some of those. I feel like a good designer kind of has the different views in their heads, but implementing it, things turn out totally different.

Jean: It's always super time intensive and very manual to test different browser sizes. Testing in the front-end in general is a huge pain.  We're too small to have a QA team. Sometimes we'll do an all-team bug bash where we all just use the front-end. I am really excited about this centralized front-end testing. I've never seen anything that has so many of the front-end gotchas all in one place. I was looking at the quotes on the website where people said, "Look, this improved my productivity more than any other front-end tool I used." And again, we don't have any dedicated front-end people on our team, so maybe front-end people have a whole thing they do, but it seems like this automates and makes a lot of that less manual.  I'm super excited.

David: It's a really streamlined way of doing testing across multiple screen sizes. It is running multiple Chrome instances, so I found it a little CPU intensive even on my M1 Macbook. I thought that it made a lot of sense, particularly being able to inspect elements simultaneously across panes. That was useful because you can see how things are changing as the resolution is different. Like you said, the chaos functionality where you can randomly drop out network access or inject CSS and disable images, the accessibility assessments in it were pretty useful as well.

The thing that stood out to me is that this is Chrome. How long is Chrome going to keep its market share for?  Years ago Internet Explorer was the one that we all wanted to try and make work. I think it's good that Chromium, as an open-source project, is being adopted, but where's the competition for the web renderer?

Jean: That's a good question. A lot of our issues come from Safari ,or a lot of our front-end issues come from non-Chrome browsers. One of my questions was, how well does this technique extend to non-Chrome browsers? I love the abstraction of having multiple Chrome windows in a single pane. I love the automation opportunity of being able to test across all of your views at once. As someone who's thought a lot about automating back-end tests, it always blows my mind that front-end testing is so manual. I asked and around a bunch, people were just like, "Yeah, this is just how it is." I think the abstraction that they introduced of having all of your panes at once, test automatically across all of them, rely on automatic testing, automatic analysis, linked to your front-end, as well as your back-end, I love it. I really don't know enough about browsers to know how transferable some of the stuff is. I agree that seeing a couple of data points extending beyond Chrome would be really cool.

David: All right. Well, that's it for this week, please let us know what you think on Twitter or by email. I'm @davidmytton on Twitter and linked in the show notes. And don't forget to subscribe to the weekly Devtools newsletter at

Jean: I'm at @jeanqasaur on Twitter. You can also follow Akita software @akitasoftware for the latest in API observability, understanding your systems from the outside. And I will see you next week.

David: See you next week.

David Mytton
About the author

David Mytton is Co-founder & CEO of Console. In 2009, he founded and was CEO of Server Density, a SaaS cloud monitoring startup acquired in 2018 by edge compute and cyber security company, StackPath. He is also researching sustainable computing in the Department of Engineering Science at the University of Oxford, and has been a developer for 15+ years.

Jean Yang
About the author

Jean Yang is CEO of Akita Software. Jean earned her PhD in software correctness and programming language design from MIT and then became a professor in computer science at Carnegie Mellon University before she started Akita to build the future of API observability.

About Console

Console is the place developers go to find the best tools. Each week, our weekly newsletter picks out the most interesting tools and new releases. We keep track of everything - dev tools, devops, cloud, and APIs - so you don't have to.