Last week I went to the React Europe conference 2016 in Paris. It was a lot of fun and inspirational as well. I actually hadn't used React for about 6 months because I've been focusing on server-side stuff, in particular Morepath, but this really makes me want to go and work with it again (I'm available). I especially enjoy the creativity in the community.
In this post I want to give my impression of the conference, and highlight some talks that stood out to me. There are actually too many to highlight here: I thought the talk quality of this conference was very high. I also appreciated the wide range of topics -- not everything was about React directly. More of that, please!
I was quite worried about travel this year. I'm in the Netherlands, so it all should be so easy: hop in a train to go to Rotterdam, a 45 minute ride. Then take the Thalys that speeds from Rotterdam to Paris in about 3 hours. In total it takes about 4 hours. Awesome.
But it's strike season in France. Railway strikes were threatened. And then there was a railway strike in Belgium, through which the train passes, on the day I was to travel. Uh oh. I already got some warnings in the days in advance about possible train cancellations due to the strikes. But my train was still going.
But travel in the Netherlands at least wasn't disrupted, so I wasn't worried about that. I made it in time to the normal intercity train that brings me from my home town, Tilburg, to Rotterdam. Found a comfortable seat. All ready to go. Then an announcement: please leave the train as it can go no further. A cargo train had broken down ahead of us. Argh!
In the end I managed to get to Rotterdam and catch a later Thalys, and made it to Paris, just 2 hours later than I'd planned.
I was also worried about announced strikes in the Paris metro system on the day of the conference. Getting around in Paris is very convenient with the metro, but not if it isn't going. In the end the metro wasn't affected.
What I did not anticipate was the whole flood situation, to the point where they had to move parts of the inventory of the Louvre. But Paris is a big city and the floods did not affect me.
So in the end what I worried about never happened and stuff happened that I didn't worry about at all.
Hackathon and MobX
Like last year there was a hackathon one day ahead of the conference at the Mozilla offices in Paris.
Last year's hackathon was special: I met up with Lee Bannard and we worked on reselect, which became quite a popular little library for use with Redux. You might enjoy my story on that.
I was very happy to see Lee again at this year's hackathon. We didn't create any new code this time; we spent most of our time learning about MobX, which I first heard about that day. We met Devin Pastoor at the hackathon. He already had a little app that used MobX that he wanted to work on together. Lee and myself helped a little with it but then got distracted trying to figure out how MobX's magic works.
MobX is a state management library, typically used with React, that takes a different approach than the now dominant library for this, Redux. Mobx lets you use normal OOP style objects with state and references in your client-side model. Unlike Redux is does not require you to normalize your state. MobX observes changes to your objects automatically and is very clever about only updating the parts of the UI that are affected.
This gives MobX different tradeoffs than Redux. I haven't used MobX in practice at all, but I would say MobX is less verbose than Redux, and you get more performance out of the box automatically. I also think it would be easier for newcomers to adopt. On the other hand Redux's focus on the immutable state constraint simplifies testing and debugging, and opened up a rich ecosystem of extensions. Redux's implementation is also a lot simpler. People want a simple answer to "what is better", but these really are tradeoffs: which way is the right way for you and applications depends on who you are and what you are working on.
Sorry folks, Lee and I created no thing new this time. But we had fun.
Dan Abramov: The Redux Journey
Dan Abramov, one of my open source heroes, gave a very interesting talk where he talked about the quick and wild ride Redux has been on since last year. Redux was indeed everywhere at this conference, and people were building very cool stuff on top of it.
Dan explained how the constraints of the Redux architecture, such as reducers on immutable state, also lead to its stand-out features, such as simple debugging and persisting and sharing state. He also spoke about how Redux's well-defined minimal contracts help its extension and middleware ecosystem.
Dan's talk is interesting to anyone who is interested in framework design, even if you don't care about React or Redux at all.
Lin Clark: A cartoon guide to performance in React
Lin Clark gave a great talk where she explained why React can be fast, and how you can exploit its features to help it along. Explaining complex topics well to make them seem simple is hard, and so I appreciated how well she accomplished it.
If you are new to React this is a really good talk to watch!
Christopher Chedeau: Being Successful at Open Source
Christopher described what strategies Facebook uses when they open source stuff. I especially liked the question they made sure to ask: "What did you struggle with?". Not "what do you want?" as that can easily devolve into a wishlist discussion, but specifically asking about problems that newcomers had. One fun anecdote: the way to make a FAQ go away was not writing more documentation but changing an error message.
I also liked the clever "fake it until you make it" approach to making a community appear more active than it is in the early stages, so that it actually becomes active. One trick they used is to ask people to blog about React, then publish all those links on a regular basis.
As an individual developer who occasionally open sources stuff I must point out rule 0 for open source success is "be a huge company with lots of resources like Facebook". Without those resources it is a much bigger struggle to build an open source community. (It also doesn't help that with Morepath I picked a saturated space: Python web frameworks. It's hard to convince people it's innovative. But that was the same problem React faced when it was first released.) (UPDATE: of course Facebook-level resources are not required for open source success, there are a lot of counter examples, but it sure helps. The talk mentions a team of multiple people engaging the community through multiple routes. A single individual can't replicate that at the start.)
Nevertheless, open source success for React was by no means guaranteed, and React helped make Facebook's reputation among developers. They made the React open source community really work. Kudos.
Dan Schafer: GraphQL at Facebook
I liked Dan Schafer's talk: a nice quick recap of why GraphQL is the way it is, some clear advice on how to deal with authorization in GraphQL, then a nice discussion on how to implement efficient queries with GraphQL, and why GraphQL cache keys are the way they are. Clear, focused and pragmatic, while still going into to the why of things, and without overwhelming detail.
Jeff Morrison: A Deepdive into Flow
Cheng Lou: On the Spectrum of Abstraction
This talk, which isn't about React, really stood out for me, and from what I heard also resonated with others at the conference. It tied neatly into the themes Dan Abramov already set up in his opening talk about Redux. Dan told me later this was not actually coordinated. The ideas are just in the air, and this speaks for the thoughtfulness of the React community.
Cheng Lou's talk was a very high level talk about the benefits and the costs of abstraction. This is something I care about a lot as a developer: how do I avoid over-engineering and under-engineering (I've written about it before), and solve problems at the right level? Software has many forces on many levels pulling at it, from end-users to low-level details, and how do you balance out these forces? Engineering is so much about dealing with tradeoffs. How do you even communicate about this?
The next day I had an interesting chat with Cheng Lou about his talk, where he discussed various things he had to cut out of his talk so it wouldn't be too long. He also mentioned Up and Down the Ladder of Abstraction by Bret Victor, so that is now on my reading list.
I highly recommend this talk for anyone interested in these topics.
Preethi Kasireddy: Going from 0 to full-time software engineer in 6 months
This was a 5 minute lightning talk with a personal story: how overwhelming software development is to a newcomer and how it can nonetheless be learned. During the talk I was sitting next to someone who was relatively new to software development himself and I could see how much this talk resonated with him.
Preethi Kasireddy also encouraged more experienced developers to mentor newcomers. I've found myself that mentoring doesn't have to take a lot of time and can still be hugely appreciated. It's fun to do as well.
A new developer is often insecure as there are just so many things to grasp, and experienced developers seem to know so much. Ironically I sometimes feel insecure as an older, more experienced developer as well, when I see people like Preethi learn software development as quickly as they do. I certainly took more time to get where they are.
But I'm old enough to have gotten used to intimidatingly smart younger people too. I can keep up. The Internet overall helps with learning: the resources on the Internet for a new developer may be overwhelming, but they are also of tremendous value. Preethi called for more intermediate-level resources. I am not sure this series I wrote counts; I suspect Preethi is beyond it, but perhaps others will enjoy it.
(Video not up yet! I'll update this post when it is.)
Jonas Gebhardt: Evolving the Visual Programming Environment with React
This was another one of those non-React talks I really appreciated. It is related to React as it is both inspired by functional programming patterns and component-based design, but it's really about something else: a UI to construct programs by connecting boxes with arrows.
There are many of these around. Because these don't seem to ever enter the daily life of a programmer, I tend to be skeptical about them.
But Jonas Gebhardt acknowledged the prior art, and the approach he described is pragmatic. An open world approach in the web browser, unlike many of the "we are the world" sandbox implementations from the past. Annotated React components can serve as the building blocks. He even sketched out an idea on how to connect UI input and output to custom user interfaces in the end.
So I came away less skeptical. This approach has potential and I'd like to see more.
Bonnie Eisenman: React Native Retrospective
React Native is a potential game changer to me as it lets people like me use our deep web development experience to build phone apps. The talk made me excited to go and play with React Native, and I'm sure I wasn't the only one. In a chat afterwards, Bonnie confirmed that was a goal of her talk, so mission accomplished!
Phil Holden: subdivide and redux-swarmlog
Phil Holden gave a 5 minute lightning talk, but please give him more space next time. He discussed Subdivide, an advanced split pane layout system for React, and then also discussed another mind-blowing topic: using WebRTC to create a peer to peer network between multiple Redux frontends, so that they share actions. This lets users share data without a server being around. This he packaged as a library in a package called redux-swarmlog.
I've been thinking about peer to peer serverless web applications for some years as I believe they have the potential to change the web, and Phil's talk really reignited that interest. Peer to peer is hard, but the technology is improving. Later that day, I had the pleasure of having a brief chat with Phil about such wild topics. Thanks Phil for the inspiration!
(Video not up yet! I'll update this post when it is.)
Andrew Clark: Recomposing your React application
Andrew Clark is Internet-famous to me, as he created Flummox, the Flux state management framework I used before switching to Redux (Andrew in fact co-created Redux). In this talk he discusses recompose, a library he wrote that helps you do sophisticated things with pure, stateless function components in React. I need to play with it and see whether it fits in my React toolbox. Andrew also described the interesting techniques recompose uses to help reduce the overhead of small composed functions -- this highlights the properties you gain when you stick to the pure function constraint.
Jafar Husain: Falcor: One Model Everywhere
When multiple development teams have a similar idea at about the same time, that may be a sign the idea is a good one. This happened to me when I came up with a client-side web framework few years ago, thought I was onto something new, and then Backbone emerged, followed by many others.
Jafar Husain in this well-done talk described how Falcor and GraphQL were a similar solution to similar problems. Both Falcor and GraphQL let the client be in control of what data it demands from the server. He then highlighted the differences between Falcor and GraphQL, where he contrasted Falcor's more lightweight approach to GraphQL's more powerful but involved focus on schemas. It's tradeoffs again: which fits best depends on your use cases and team.
Laney Kuenzel & Lee Byron: GraphQL Future
This was a wide-ranging talk that went into various issues that GraphQL team at Facebook is trying to solve, mostly centered about the need to receive some form of immediate update when state on the server changes. Laney and Lee presented various solutions in a various states of readiness, from mostly untested ideas to stuff that is already deployed in production at Facebook. Very interesting in you're interested in GraphQL at all, and also if you're interested in how smart people tackle problems.
In my blog post last year I was clear I enjoyed the conference a lot, but also engaged in a little bit of constructive criticism. I don't presume that the React Europe organizers directly responded to my feedback, but let's see how they did anyway and give a bit more feedback here. My intent with this feedback is to do my bit to make a great conference even better.
Last year the conference was in early July in Paris and it was 40 degrees Celsius. The React Europe team responded by shifting conference a month earlier. It was not too hot: problem solved.
Last year the hackathon assumed people were going to compete in a contest by default instead of cooperate on cool projects. This year they were very clear that cooperation on cool projects was encouraged. Awesome!
Still, I found myself walking around Paris with a friend on Friday night trying to find a quiet place so we could look at some code together. We enjoyed the conversation but we didn't find such a place in the end.
This is why I prefer the approach Python conferences take: a 1-3 day optional sprint for people to participate in after the conference has ended. Why I like afterwards better:
- You can get involved in cool projects you learned about during the conference.
- You can get to know people you met during the conference better.
- Since there is no pressure it's a good way to wind down. Speakers can participate without stressing out about a talk they will be giving soon.
Many of the speakers at this conference work for Facebook. They gave excellent talks: thank you. I understand that having a lot of speakers from Facebook is natural for a conference on React, as that's where it originated. (and Facebook hires people from the community). But this is an open source community. While I realize you'd take on more unknown quantities and it would be more difficult to keep up the quality of the talks, I would personally enjoy hearing a few more voices not from Facebook next year.
Last year I spoke about a bit gender diversity at the conference. This year there were more female speakers than last year (keep it up!), but male voices were still the vast majority. Women speakers are important in helping women participants feel more welcome in our conferences and our community. We can still do a lot better: let's learn from PyCon US.
The train ride back home on Saturday morning was as it should: uneventful. Left the hotel around 9 am the morning, was back home around 2:30 pm. I came home tired but inspired, as it should be after a good conference. Thanks so much to the organizers and speakers for the experience! I hope you have enjoyed my little contribution.