Few seats left for our Advanced Backstage Plugin Development workshop on December 7th. Secure your seat today!

109: What Do You Need in a JavaScript Framework?

Hosted byCharles Lowell and Taras Mankovski

August 23rd, 2018.


In this episode, former Fronsiders, Brandon Hays and Chris Freeman join Charles and Taras to talk about the difference between a framework and a library, whether or not React + Redux a framework in itself, red flags to signal that you’re actually building a framework, attributes of a good framework, how can you tell if you created a bad framework, and how you can make a bad framework better.


This show was produced by Mandy Moore, aka @therubyrep of DevReps, LLC.

** Transcript:**

CHARLES: Hello everybody and welcome to The Frontside Podcast. My name is Charles. I'm a developer here at Frontside and today, we're going to be talking about the things that go into making a JavaScript framework. Because, hey, there's not enough of those in the world today, so we're going to talk about that and with me is Taras.

TARAS: Hello, hello.

CHARLES: And we've got two very special guests, who have a lot of experience with this topic. Mr Chris Freeman and Brandon Hays. Hey, guys.

CHRIS: Hi, there.

BRANDON: Hi, there. We're talking about the poofberry framework, right?

CHARLES: What's a poofberry?

BRANDON: There's a tweet that's going around right now that one of them says, "I don't know what I should be doing," and the next person says, "Oh, just use poofberry." What is that? It's like fluffnuts but the [inaudible]. Hey, dot, dot, dot. Then, it integrates with log bungler.

CHARLES: There's a reason that I'm dying laughing.

BRANDON: It's so true.

CHARLES: It's so true, laugh, cry, laugh, cry. Let's start with kind of a very basic assessment here. Because there's a lot of different things that you can use to compose the applications that you build but for some reason, some of these things are grouped and considered as libraries and some of them are considered frameworks. I don't know that the boundary is very clear like I'll know it when I see it type thing. Maybe, we can start with what is the difference between a framework and a library?

CHRIS: I have some thoughts of these. I feel like this is one of those questions that could easily just turn into an infinite bike-shed but I remember reading something a while ago that stuck with me for a long time. I'm pretty sure it's related to Java but that makes sense because if anyone is going to talking about frameworks, it's Java developers. But it was saying that the difference between a library and a framework is inversion of control and the idea is a thing that's a library is a thing where you are in control. You bring the library code into your code and it's up to you what you do with it.

In a framework, the framework code calls you as I think what it said. It's like, you call the library code, the framework code calls you and --

CHARLES: In Soviet framework.

CHRIS: Yeah, exactly. A framework says, "Here are a bunch of open spaces for you to put your code in and I will take care of the rest," versus a library is just like... I don't know, "Here's some things that you can use. It's up to you. What do you want to do with them?"

CHARLES: Right, so in kind of like [inaudible], that would be basically, a framework would be the thing that's got the main method. I think the same thing in JavaScript and when you call it, does it actually implement the main method. In JavaScript, it'll probably like in node. Under that definition, it would be like, "Are you the main scripts when you invoke node? Do you control the main script?"

If you were doing your own command line parsing, for example, you're looking at the process.rb and pulling off the command lines and doing all the things but even if you're using something like Yargs or option parser in Ruby, that's more of like a framework. I guess Yargs is a library because you're still implementing the script. You're instantiating the Yargs thing.

TARAS: React calls render to figure out what to convert to DOM. Does that make React the framework?

CHARLES: I think React as a library. That's a good question. What's the equivalent of the main method on the web?

CHRIS: I think there's a very clear distinction, especially if you look at React versus something like Ember and I'm sure Angular does this as well. In React, by default, to build a React thing, you're going to pull in React, you may write some components, you may import them elsewhere but the main method is that you have an index.html with some div in it and you are the one that has to call ReactDOM.render and you pass it like document.query selector or whatever and then, your top level component and that can be as simple as complicated as you like or you can have a webpack plugin do it or whatever else. But the onus is on you to actually take that React app and get it starting up on the page versus Ember, it's like, "There's an index.html. It's fully wired up." There is one point where you sit down and say, "Start my program here," like Ember abstracted all that away. To me, that's the main method for a frontend application.

CHARLES: Right and if you actually look at something that Ember generates, then look at index.html, they generate a script tag for you that instantiate your application and mounts it on an element. If you want to change that element, that's actually a configuration option that you can change but it still a configuration option that's consumed by the framework. In that sense, there is that inversion of control. I see what you mean like in React, you're the one who flicks off the first domino, like who's the prime mover. Is it you or is it the framework that knocks over the first domino?

BRANDON: I like Chris's explanation and I think it's elegant to say because I was thinking in terms of structure. If it imposes a structure on you but really, the structure is there, it's like one of those Ikea shelf systems for you to put stuff into. If you're trying to solve a problem, here's a shelving system for you to put stuff into, whereas a library is just the tool that you might get out to put something together. Something that's multi-purpose but doesn't impose any structure on you or a ton of structure on you. My question is what's the usefulness of distinguishing between the two?

TARAS: I think what's interesting and I had experienced this in a last couple of projects is that people, especially React when they kind of assume, because a lot of people entering to React not understanding the context within which React emerged and so, they're getting into React assuming that it has everything you need to build application that you need to build. A lot of them haven't necessarily built a single page application from scratch before and so, the jump into building something with React and then, it takes about a year for them to realize the full scope of all of the features that their application actually has and then, they kind of take a retroactive look and look like, "Okay, what do I have now?" and what emerges is that they've actually over the last year, they may be creating a framework without realizing that this is actually happening.

CHARLES: They've imposed the structure of saying, "Here's the shelving system. Books about geography go here. Books on English literature go there and so on and so forth."

BRANDON: But when you rolled your own framework, that's not how it goes. It's like, "You have to launch this balloon into the stratosphere to put a book on the shelf from geology." Taras, to your point, it sounds like the importance is setting expectations properly for people, so that they know what they're in for because kind of calling back to Ryan Florence's post a few years ago, you can't not have a framework. At some point, you will have a framework in order to ship something.

I would actually take it one step further. My friend, Kyle talks about this that library is the smallest unit that you're working within a framework but that still doesn't take your code to production and put it in a debugable state. You need a platform. It's arguable, if you're handling deployment tasks and debugging tasks and operating software in production, you now have a platform and it's fair to say that Rails crossed that threshold at one point. It's fair to say that Ember has probably crossed that threshold, if you combine Ember with CLI deploy and the CLI tooling and all of that stuff. This almost like acts as a platform if you're owning and maintaining the software in production.

CHARLES: Now, can I play devil's advocate here and say, the platform, is that necessarily predicated on a framework? Is there a pyramid where it goes library, framework, platform and one is built on top of the other? Why couldn't I have a library? Because what I'm hearing is the scope of concerns is just rendering HTML based on a state is a very small chunk. The actual scope of things that you need to do to get that code in production and have it be reliable and do all of the features that you want to do is just massive but why is that predicated on a framework?

For example, one thing you have is a bunch of libraries out there, like routing for managing the title tag, managing all these things that you have to do for managing deployment, for building your application, for compressing it. There's all these different libraries out there. What if there was one massive library that just picked a bunch of other libraries but I was still in control?

TARAS: I've actually seen this happen in the last of the projects. When people jump into building, they will eventually realize that they're building a platform but what happens before that is that they take user's requirements and they break those up by sections and then they assign them to a bunch of development teams who go and actually start to build. On one platform, they end up building five or six or 10 of siloed, packaged applications that have, in some cases, have their own dependencies, they have similar architecture, might not have similar architecture.

Each team kind of implements thing differently and there's an expectation that once you package these things as npm and then you install them into one package, to one application when you run build, it's just going to work together. That's where I think, with the framework, it does create a foundation for these verticals to be implemented using kind of common foundation. This is what a lot of times that as if you don't realize that what you're trying to set out to build, the way that the projects get managed quite often, especially for big applications, for big platforms is that, it creates this period of about two years, where there's a lot of confusion and there's a lot of duplication and then, you end up seeing code that it's hard to put in production.

CHARLES: Yeah, I agree. I'm curious then, because we'd started out talking about library and framework and talked about it takes two years to recognize that you're building a framework or you're building not a framework but a platform. Brandon, you said something very interesting. Rails for example, crossed the threshold of being more than a framework and actually, being a platform. What are the concerns of a platform that are beyond a framework? We talked about and using the kind of loose definition of a framework as being something where the framework create spaces for your code, to run your code so you can just take little dollops of code and they have one concern but the framework manages the coordination of the concerns but what's the next level?

BRANDON: For the purposes of this conversation, I may have muddied the waters a little bit because I think it's more interesting to talk about the transition and the level of which you've crossed the threshold from being a library or using libraries or collecting libraries, into maintaining a framework because it's where you're going to experience more pain more than likely, than to me, the idea from works on my machine, to deployed and supported across a lot of users, it sounds like it's more interesting but it's not where we experience most of our pain actually. From my experience, maintaining frontend single page applications most of the pain is actually getting the damn thing to work on your machine and getting the libraries to collectively work together and then, getting that to production, it kind of enters back into an area of more known unknowns.

I think that's a surprisingly a more mature ecosystem, still getting from this thing works on my machine to getting it out the door. That wasn't true when Rails was invented and so, Rails had to invent a lot of its own ecosystem around this stuff. Like I said, I don't want to muddy the waters too much.

I think to me, the interesting question is how do you know you've crossed this threshold? What pain points are exposed when you start crossing that threshold or when you're pushing the boundaries of that threshold? Because you should not be using a framework if you're using React to do a select dropdown. I think of it as, if you're using it the way to replace something you might do with a jQuery plugin five or 10 years ago, you're using React like it's a library. One of the questions that you brought up was is the combination of React and Redux is a framework and I would argue that it is but I kind of want to throw that out --

CHRIS: Oh, interesting.

CHARLES: I would say, it's two libraries stuck together to make a bigger library. It's like a monolithic library.

BRANDON: But by the time you're actually using that to do anything, maybe the third thing in there is like transitioning states when you transition routes. At what point is that threshold crossed? I didn't build most of the software that led me to some of the opinions that I have about this. This was actually Chris Freeman's, though. I may defer to you on this.

CHRIS: I think React + Redux constitutes if you look at what it does. You have like this view layer and this state layer. There's a set of opinions on there that is useful and there is the foundation for doing quite a bit but in my experience, you've already kind of alluded to this a little bit. I don't think it's a framework because as soon as you start using those two things, suddenly the next thing you hit is, "Wait. How do I handle asynchronous things?" There's a lot of different options for that. "Oh, now, I need to do routing. How do I incorporate routing into my React app but also in a way, that is amenable to state transitions in Redux but also, that is aware of the async stuff that I'm doing, that is going to possibly be triggered by my routes and by my Redux actions or by some other side of things?"

Suddenly, you are very quickly pulling out a bunch of other libraries but also, probably starting to build abstractions on top of them because you're already finding a lot of common patterns that you're repeating over and over as you incorporate more and more pieces of the stack and then, you're writing a lot of glue code. I think that's the point where suddenly, you look back and behind you is the footsteps of this framework that's been walking alongside you the whole time.

BRANDON: That is where I carried you, then dropped you, then sort of drowned you.

CHRIS: Yeah.

CHARLES: And then, kicked your core.

TARAS: I'd like to suggest a way to think about this. As you guys are talking about, it kind of occurred to me is that it seems to me that libraries concentrate on how and frameworks focus on the 'what.'

CHRIS: Oh, I love that.

TARAS: Because if you think about for example, React is how geostack efficiently update DOM, then Redux is how do you wire together state across multiple components that might be in different parts of state tree and if you look at, for example, a React router or a kind of a routing component is how do you choose which components you want to render when you navigates specific URL. Because those things by themselves are not a complete solution but when you combine them together, what you get is you have a way of saying, "When I navigate to specific URL, I'm going to load specific data, provide that data to components and then, I would have a way to navigate through a different URL when you click on a link."

From that, I think what happens when you get to the framework level is you actually have a kind of a bigger umbrella and under that umbrella, you have ways to address problems that you did not have previously. I think that's what framework does it is over time, it's a way of addressing concerns that cannot be addressed with a solution. They have to address with a collection of solutions and then, they provide a specific solution. I don't know if that's --

CHARLES: That actually sparked off a train of thought in my mind that perhaps what you really want to do is say, "I'm going to go a little bit like Lisp on you all," in the sense of every code at some point is data, that maybe every library, at some point is a framework. It's just that you can look and say, "What is the scope of the 'what' that I'm tackling." For some point, you can say like React is a framework. It creates this space where I can put my JSx, AKA the render function and I'm basically inverting control and so, what it is, it is a framework for efficiently rendering HTML or efficiently mapping an object to a fragment of DOM and then the DOM that gets generated from your render function, patching that into the HTML. You don't have to worry about that. There's that inversion of control. It creates that space but that's the only space that it creates.

From that perspective, React is a framework for generating HTML but that's all it is but it is a library for constructing applications. Does that make any sense? I think as you layer on concerns, your framework create spaces for you. You use your library code to put stuff in and so, in the same way, I think one of the key realizations, I'm going to call up like BigTest and I'm not going to take credit for this, which is actually a blog post that I read at Google. I can't remember what it is but we'll link to it in the show notes where he said, "There are no such thing as unit tests. There are no such thing as acceptance tests. There are just tests of varying scope." They're all acceptance tests. To use that one thing, they're all experiments. It's just what is the scope of the test that you're trying to accomplish and his argument was we want to make that scope as big as possible by default and then, where appropriate, you narrow down. Maybe, the framework library distinction is a little bit constructed, kind of a construction of our own minds and what really is there, there's just frameworks of varying scope.

BRANDON: Agreeing on a shared scope is actually probably the most important part of this conversation. We're referring to building end-to-end an application from data access to rendering to testing --

CHARLES: To deployment to routing.


CHARLES: To one day accessibility.

BRANDON: Yeah. Adding that into the discussion is like a baseline of what constitutes an application. It's the percentage of people that are able to actually use it, the people that are locked out from using it by ability. That's a very useful frame for the discussion. Let's agree on the scope of what an application is and then, coming back to what Taras was saying is basically, when you're talking about the 'how,' that's a decision point.

You hear a lot of people talk about decision fatigue in JavaScript and it's almost a played out trope at this point but it hasn't gone away as a problem, so what frameworks are doing is they're making a series of decisions for you that allow you to basically connect the pieces from end to end. Basically, somebody threw a rope bridge across the canyon and it doesn't have to be the best solution to get end to end but we have to solve the problem end to end.

If we agree on the places end to end and the problem is when you're building your own series of libraries, you're like, "I'm going to choose best in class of A, best in class of B, best in class of C," and that sounds really good but if you're trying to build a bridge across a canyon and you're building in 10 best of class sections, for the type of connection we're trying to make here in the middle, we're going to use the best in class here. The weak point is in the connections, so you had better be the world's foremost engineer if you're going to be the person connecting all these disparate pieces that are each best in class, in order to bridge this canyon.

That's the thing that's interesting to me and it's not even agreed in our industry that JavaScript-based web applications are a good thing or that the browser is web application runtime, those are things that are up for debate. But I think if we make that assumption, this is sort of the founding principle of where Ember came from and it executed to the best of its ability at the time and that philosophy is, I think you can prove it out in terms of results based on if you have two different applications, one of them is built by somebody trying to jam together best in class components and the other person is starting with an end to end solution with a community of people rallied around that solution. It's been interesting to watch those approaches play out over time. I know Chris has a very specific hands-on experience of having done both of these. I'm curious to get your hot take.

CHRIS: There's actually a concept that I think about a lot in relation to this question. It's something that I actually heard come up again recently so the timing was great but it's called hypocognition. The idea is hypocognition is when you either just like can't see or can't understand some kind of cognitive representation of something because you don't have the words for it. An example is in Western cultures, especially like in English speaking cultures, there are not that many words for the color blue but in a lot of other cultures, they have many, many words for the color blue. After doing a big study they found that these English speakers actually have a harder time recognizing different shades of blue, like more of them just look the same versus other cultures where their brains are actually wired to see all this variety because they actually have the linguistic representations for these ideas already.

When you were talking about maybe a library is a framework at some point, I think that's right on. I think one of the things that I think about a lot when talking about frameworks and seeing these debates happen on the internet about, "What is a framework?" but also like, "Do you even need a framework?" is obviously, there's a lot of people who absolutely... Like Ryan Florence. Ryan Florence clearly knows what a framework is. He knows what it takes to build a web application and he does not lack the words to define a framework versus a library. He's just made that choice and it's a very informed choice but I wonder if there's also a lot of people who are getting into web development for the first time and they look at something like a framework and it seems just absurd to anyone would want all of the things that like in Ember or in Angular is talking about, when they can make a basic UI with React and it's easy and fun and really cool.

But then this two-year path happens and they look back and they've learned a whole bunch and now it's like, "Ooh, you couldn't even have explained this to me before," because all of the words would have fallen on deaf ears but now suddenly, it makes a staggering amount of sense.

CHARLES: Right. I love that.

BRANDON: You have to make a bad one.

CHARLES: Just so that you can inherit the vocabulary to understand why you made a bad one. Now, you guys actually have some experience with this. Brandon, you gave a talk about it, which I think you should give more widely because it's fantastic but for those folks who may or may not be aware that they are walking this to your path, I want to talk first about what are the signs that you're walking along this path and then two, what are the consequences in terms of the cost you're paying for walking this path. Let's start with that first thing. What are the signs? How can you tell that I am building a framework?

CHRIS: I think one of the telltale signs and one of the biggest red flags that caused me and Brandon to have a very serious heart to heart about our own personal framework was when we hit the point where you could look at a set of tickets for features and all you saw was 'framework features' that you needed to write before you could build the feature itself. You know like, "Oh, we have basic routing setting and we have it set up so that if you have a route transition and you would like a data request to happen when a certain route transition happens, that will happen," but then someone would like infinite scroll and we want to use a query param. When a query param changes, I want to update the query and fetch more records, except that the glue code that we wrote to tie our router to our redux async stuff is not aware of query params. It has no concept of what a query param is or what to do when it changes. Also, it has no concept of refetching the data without a full route transition, so what do we do, this person wants infinite scroll but I first have to implement several layers of framework code before I build the UI feature that you want?

CHARLES: The basic heuristic there is ratio of direct feature code to code that supports the direct feature code and code that supports the code that supports direct feature code. It's anytime you're anywhere above that first layer on the stack.

CHRIS: Yeah, I think Taras nailed it like what's the 'what' versus the 'how.' If you're asked a question that is concerned with the 'what' and you spend more time focused on the 'how,' then you might have a framework.

BRANDON: I think people will think of building an application like a recipe. If you think of it in those terms, people think of frameworks as very restrictive but I'm a big fan of Blue Apron, a sponsor of this podcast. Thank you.

They pre-select the ingredients and they give you the instructions and you know what to do, you still have to do the effort but you know if you connect these pieces together properly that you're going to wind up having a good experience and then, it gives you a lot of freedom to experiment and be creative beyond that, should you choose to. I think one of the signs that you've done a crummy job is that you're staring down, like Chris Freeman said, you actually starting to restrict your choices like, "I can't actually build you that feature because we don't have time to take on the amount of work necessary to build the support structure, to build you that feature," or if you find yourself writing a test framework.

CHRIS: Oh, yeah we did that too.

BRANDON: You know, we were real deep in this. There are developers that are like, "I really want to feel like I'm walking into a grocery store and selecting all the things necessary for my recipe," and so it really depends on what the problem actually is. If you're working at a giant megacorp and you have a two-year timeline to deliver something and their goals are not about delivering stuff on a tight turnaround, that's usually a recipe for a software failure anyway but let's say, that you're in the 5% of those types of projects that's going to succeed, that might be a good place where you can say, "What we're trying to do here is so custom and we have such a long lead time and a long leash and such a high level of internal expertise here that we should be shopping in the grocery store and we should be selecting all these things and we should be solving these problems."

Basically, when is it time to use a framework? Well, when you don't have 10 times the time you think you do, when you don't have the ability to spend 80% to 90% of your time in the first three to four months of your project, maybe six months, debugging you're glue code in between the different libraries that you're gluing together and then coming back and realizing that you've painted yourself into a corner and you have to re-architect your whole framework, then you could be so proud of this baby, 18 months to two years from now, when you actually have delivered both a framework that took about 70% of your time and an application that took 25% or 30% of your time.

CHARLES: Yeah. I think it's important to realize that people think we'll do it and we'll build it as we go but I want to call out right there, you will be spending 80% of your time and you have to be upfront about it. Of this two years, 18 months of it is going to be spent building this framework and six months of it is going to be spent actually writing the feature code and you have to be 75% of your tickets or your issues, whoever track the work, 75% of that has to be dedicated to the framework.

BRANDON: If you're going to bake in that kind of overhead purely for the satisfaction of a single or one or two developers that like inventing things, that is literally the worst possible reason you could do that. That is almost like a guaranteed recipe for failure. It has to be for some other business reason like, "We want to be the company that owns this." There has to be business value attached in making that kind of investment. If you can't justify that at the outset, then you should probably just go ahead and lean on an existing framework and join a community of people.

CHARLES: Yeah and I think one good litmus test for that is, "Is this a 'what' for which there is currently no 'how?' One of the reasons we're writing BigTest is because for the general JavaScript community, there are a number of acceptance test frameworks out there but the market is very, very limited. When we look to actually acceptance tests, our React application, this thing does not exist. Now, we had experience with something that was very like Ember specific and so, we kind of knew what the 'what' was, we experienced the 'what' but there was no 'how' for our current situation. That's like a place where you might be called upon where makes business sense to actually invest in a framework.

I'll tell you another thing too is if you have made the decision to kind of follow the beaten path on the other areas, then when a framework is called for, you have the bandwidth. You've allowed for the buffer, for the margin, for you to write in with that framework, whereas if you're already just by default, maintaining all the glue code in every single thing, then if some unique 'what' comes along, for which there is no 'how,' you're not going to have a bandwidth to tackle it.

BRANDON: Yeah. That's a real bad situation to be in.

TARAS: There's something else that I find interesting is because there's a certain point, like this two-year mark where everyone's like, "We want to fix this now." I think what is interesting what comes next which is the three years of undoing all the stuff that you made because the biggest challenge, especially in really big projects. When your projects has to borderline into platforms and a platform threshold is when you have a multiple teams working separately to write separate modules that run, maybe in a separate Git repo and maybe, packaged in separate npm package and assembled together.

Then what happens at that point, the question arises like how do you actually make this changes in this environment. Answering that question is actually really difficult. I think if you look at frameworks like Ember, Ember has made it their business to figure out exactly how to make this happen and I think they've done it really well but it's a really challenging endeavor, especially in incorporate environments where they don't have an update. You have like upgrades are like a curse. It's like a thing that you don't really want to ever do and because most quite often, they don't have the right testing habits in place to be able to support the change if necessary.

I think what a lot of times happens is that the team that made the framework in the first place, they end up trying to maintain a fort but you won't have like 10 people and they only have machetes, you know? All you can do is run around and try to chop down little twigs but at the end of day, the trees is still going to keep growing. I think that's the really challenging part of being two years into a project, where you realize that you actually need something much more comprehensive than initially thought you needed.

CHARLES: On that, assuming that you have decided that you are going to make a framework, it's a good business decision for you. Based on the criteria of this discussion, how can you assess whether it's good? Chris, you talked about needing to integrate query params with routing and asynchronous data loading and making sure all of that coordination happened and worked together easily. What's the difference between your framework just missing features kind of having holes in it that can be filled in, versus something that's not good and it's going to cost you lots of money down the road.

CHRIS: Yeah.

TARAS: One thing, if you look at what makes a good library of any kind, it tends to be like how effectively and how much words to take the address the use cases that you need. The problem is that to build a good framework, you need to understand the use cases. This is what usually happens over time. Two years in, you've actually understood the use cases and now, it's time to change and so, I think if you want to build a good framework, you actually need to understand those use cases quite early on or account for understanding use cases over time and that's a big question -- how do you figure out how to know what you don't know.

CHRIS: Yeah, I think that's exactly right. I think about what you were just saying Charles and Taras like one of the things that I think has a big impact on and what this process looks like is the completeness of vision for what's your project actually is. If you have a very, very clear idea of what the entire product you're building is going to be or, at least what the key money-making feature is going to be and you can understand the ins and outs of that, then I think that's the point where you can look at what you have and say, "Have I created a good or bad framework? Does this framework have the ability to solve this one very important thing that I have to be able to do? If the framework doesn't do it, then I need to build my own but I now know what very important features I need to front load my framework with."

I kind of think of it as imagine that you're like Jeremiah Johnson, the Reverend Jeremiah Johnson and you're going to go trekking through the woods for some unknown amount of time and you have no idea yet. You don't actually know where you're going. You don't know what you're going to see. You don't even know what's out there because you haven't done the research or whatever and you need to be prepared for anything, so you bring just a hodgepodge of stuff.

If that's you at the beginning of your company or the beginning of your product and yours is kind of like... I don't know, we got to get product market fit and that means that we may have to kind of pivot once or twice or we need to be very flexible, then I would think long and hard before you commit to writing your own framework because you don't even know what framework to build and you might as well take a broad array of tools and use what you need.

There will be times where that's frustrating and there won't be exactly the right tool for the job but 80% of the time, it's going to do just fine but if you know you have to do this one very special thing and you know that a framework is going to give you a lot of stuff that you won't need and it doesn't really excel at the one thing you do need, then don't force the framework. There may be time to build your own but just know that you need to go in with a very clear idea of what you're doing before you start building the abstractions that constitute a framework, rather than just like a constellation of libraries.

CHARLES: I have a question on that then. Going back to one of the things we were talking about like React plus Redux. Your opinion, Chris that it is not a framework, so the question is does a framework actually exist for React?

CHRIS: My guess is that many frameworks exist for React.

CHARLES: Is there a public framework?

TARAS: There is one called Fusion but it's [inaudible] what you would have imagine. It is essentially Redux and React together conventionalized. They addressed a bunch of concerns around service rendering and such but it does exist.

CHRIS: How about Next? Next.js?

TARAS: I'm not familiar with its features from a single page application perspective.

CHARLES: I think it does have a router. It does bundle with Redux and this is one of the things that when you first started using Redux, it's like, "How do I even get my store to my components?" Yes, I can connect them but there's actually a lot of stuff that you have to do. First, you have to say, "I'm going to put my reducers here and then when I create my store, I'm going to fold all my reducers. If I've got a whole bunch of reducers in my application, I've got to fold them all together. I've got to pass them off to the store. When I create the store, I have to inject the middleware and then, everybody else just imports my store and then, I have to put in a provider and then, I can connect my components."

That's actually a lot of stuff that you have to do and I think that, for example, Fusion just says, "Put your reducers here and we'll take care of all that process," and so it makes that decision for you, right? It says, "For state management, you're going to use Redux. For your reducers, they're going to go here. For your actions, they're going to go here." I don't know exactly how it's laid out but I remember reading the ReadMe, it was basically layering conventions over that. That's definitely going into framework territory but that's the only one that I know of, which is really, really odd.

TARAS: There's something interesting that's happening also and this goes to what Brandon was saying earlier is that choosing the best in class, there's this 10 things but then, what if one of the best in class stops being the best in class. The fact that the creators of Redux was essentially saying that we needed to basically provide a way to do Flux that was better than 10 different options that were available, so here's Redux. We've created Redux but we don't really think it's ultimately the solution. We need to have something else in React that provides a foundation for us to be able to deliver a better state management than what Redux is, so what happens when one of the best in class is no longer the best in class? The bridge is already standing. There's people walking across the bridge already. How do you replace one of the chains in it?

CHRIS: Over the course of six months while you figure out the differences in API between Flux and Redux and all the custom route transition data loading stuff you did with your Ajax library in your state management software that you put in a case statement inside there that you now have to change over. It's easy. It's no big deal. Don't worry about it.

BRANDON: Just a simple matter of programming.

TARAS: At least 25 years of collective frontend development experience is laughing like hyenas about the simplicity of building a --

BRANDON: Yeah, I'm actually looking at some of the old code that Chris wrote for trying to glue together, Redux Saga. I've been out of the game long enough to not know whether that's been superseded by some new or best in class piece of technology and even then, it was really challenging. This is true for frameworks too, is they don't really optimize for best in class. They optimize, hopefully for best fit for purpose but the world has moved on since Ember launched obviously. A lot of things have changed and it's, at least as difficult to try to keep that up to date with evolving trends and technologies and updates for a core team at a framework level as it is for you, as an engineer on the team. The difference is you get to outsource that work to a core team for a framework.

Ember has not done a fantastic job in keeping up with. They've done a good job and they've tried their best but if there were more people working on it or if there was more effort applied to it or if it was a higher priority, you would see Ember being a more up-to-date framework using more modern tools. As a framework author, if you stay too close to the bleeding edge, all you're going to do is change out your build system. You're going to replace a Broccoli with webpack, with Rollup, whatever's after that. What's new in Packer?

CHRIS: Parcel?

CHARLES: Parcel.

BRANDON: Parcel. You should immediately go build your framework with that and have fun. I am excited by the new and interesting stuff that's happening in these ecosystems and I think it's important not to get lulled into the siren song if your goal is to actually ship a piece of software on a timetable or a budget.

TARAS: One thing, if it's a red flag, if you think this is easy, if you think your decisions can be made in this isolation without talking to somebody else and actually kind of flashing it out, then you're probably doing something wrong because a lot of these things are not trivial. There's a lot of thought, there's a lot of considerations used to go into decisions that you make, especially when you're creating something that is going to be used by more than a few people. I think that's really one of those things where it's hard to know what you don't know but if you think you know and you haven't done this before, you haven't done this a few times before, you're probably missing some pieces.

BRANDON: Yeah, I agree with that.

CHRIS: I think one of the things that's really enticing about React and Taras, you just hit on it but I've never felt as clever as when I was writing a React app. If I'm clever, I mean, clever in the same way that I felt really clever when I wrote some unbelievably convoluted Scala one-liner that six months later, neither me nor anyone else could decipher what it meant but at that time, I felt like a god of programming. That's how it felt like, "Well, a lot of the React stuff is addicting." It felt so much fun. It was so much fun until I really had to do something and it mattered for my job and there was a deadline and people were depending on me and I've realized that the clever thing I had done a month later was not the right clever thing but I can see how, if you're like what Taras was saying, where you are at the point where these decisions are easy. These decisions make sense. We're going to be fine and you haven't done it enough to kind of like know where all of the pitfalls are. That cleverness that you feel is fantastic and I can see why it takes two years before you look back and if the cleverness was finally worn off and then, you're just mortified at what you've done.

CHARLES: Pride cometh before the fall.

CHRIS: Yeah.

BRANDON: It's like being a dungeon master in Dungeons and Dragons, where you're like, "Oh, look at this fiendish world." All right, cool. Now, you actually live there though. I have to move into an apartment on Mordor.

TARAS: You know what's the funny flipside to that is that coming from Ember world where it's so normal to leverage the work of other clever people, like really smart people who've invested a lot of time to solve a particular problem, is that there's no stronger sense of being dumb than having to write it from scratch in React. That first feeling of like, "I've actually never had to implement this from scratch," and I feel like a bunch of applications before but because I've leaned on for accessibility, I've leaned on something that someone else has done and it worked really well for me and it was perfect. But now, I need to implement autocomplete from scratch in React and I have no support. I'm basically learning as I'm going on this and it's that sense of discomfort that you get from having to do it from scratch and then, comes the euphoria of having to figure it out. But if you figured it out, you figured it out in the last month. You've written it for the first time in the last month and you now understand what all the things that the Ember implementation does for you. It's an interesting psychology of doing this --

CHARLES: Yeah, it gives you a lot of perspective but you have to ask as a business owner, who may or may not be technical and this is the hardest thing for technical people who are business owners is to be able to not see things through a tactical lens. Is what you really want to pay for is to basically give your programmers this kind of a-ha moment of their own shortcomings because that what you want to be buying.

BRANDON: Yeah, you want to maximize leverage. Your goal with technology is to maximize leverage. It's like being hired as a chef and you walk in and then you're like, "I'm a terrific chef. I worked in these fancy kitchens in New York and I'm known as a great chef," and they're like, "Okay, cool. Here's some flint and steel and a spear."

CHARLES: Go hunt.

BRANDON: You're like, "Wait, what?" Yeah, yeah, yeah. Show me what you can do.

TARAS: We had a conversation in one of the previous podcast with Michael Jackson and we asked him, "What is the one thing you wish like React community would do more of?" and he's like, "I really wish React community have more conventions." All of this is to kind of say as like, there is a place for frameworks in React world. There's a very strong place for it. The question is how and what it said and how do you actually build it and when do you --?

BRANDON: So we need a framework for making framework.

TARAS: Getting really meta here.

BRANDON: I totally agree with that and that's a great observation and that was actually the point of my talk as well, which is if I could convince people just to use Ember and improve Ember, that would be great because I think it's a really great starting point. But the React community is much larger because it had such a great adoption story. The adoption of Ember was very difficult and the adoption of React was very easy and it expanded to include the scope of full end to end applications in terms of what people thought the problem spaces they were thinking of with React.

Ember was built to solve that but it was hard to get into. React was really easy to get into but it's actually hard to build applications with. I would love to see a dedicated subset of the React community, except the idea of shared solutions and the philosophies that made Ember into sort of a powerhouse of value delivery but built out of tools that satisfy the React community and a little more modular and a little more available for people to customize and built in that ecosystem.

I'd really love to see that that included all of the main components of what we accept as, "This is an application framework. It handles testing. It handles accessibility. It handles data loading," and it doesn't have to be best in class in every scenario but it does have to be a reasonable bridge across that chasm and have a group of people look at this the same way. I would love to see a collective subset of the React community dedicate themselves to this idea. I don't know if that's too culturally opposed or even orthogonal to what the value system inside the React community. I haven't been able to fish that out but I would really love to see that emerge.

this is something I would love to push for and I'd love to see other people jump in and push for as like, "What if 20 of us got together and decided we're all building our applications in similar ways, instead of one person saying, 'I'm going to use --'" Even create React app is kind of a Band-Aid on that, it isn't useful past a certain stage of life. I would love to see a group of people, though, get together that are sort of like-minded like that, the Michael Jacksons and maybe even Dan Abramov or a group of people that shared that set of values or came into React from the Ember community.

That’s actually one piece of advice I would give to people. You said, "How do you convince this engineer that they've built a bad framework?" Use a decent one. That's the biggest guide. Use a decent one. Build something in Ember and ship it to production and go, "Oh, I get it." If you've used a good framework, you can't go back to rolling a crappy one. Your standards have been ratcheted up.

CHRIS: I wholeheartedly agree that you should try something else and Ember is a great option but I don't want to dismiss just like, "React is cool as hell," and there's a lot of stuff in React that's really, really awesome and things that I wish that will show up in Ember and they are starting to show up in Ember but they're taking a while and it'll be nice in there but who knows when that will be but I would encourage even more so is both sides, like Ember folks who are listening to this podcast, if you have never messed around with React because you feel some kind of tribal affiliation that you can't betray, please set that aside and go do something in React because you will learn a lot about why Ember does what it does and you will see a lot of really interesting things that will probably jostle some ideas loose in your brain.

The same thing goes for React developers. You, 100% should spend a weekend building something in Ember and nothing about that means that you have to switch or it's going to change the path that you're going on at work but I guarantee you, you will go back to your React application with some new and fair useful perspective that you didn't have before and that's okay. That's great. There's no identity crisis that will come about as a result of that.

CHARLES: That is a fantastic advice, Chris. It will only stretch you.

CHRIS: Yeah.

BRANDON: I think developers have been sold this idea of a competitive landscape by authors of these frameworks because it helps sell the framework. You can build and strengthen a community by leaning into the tribalism that can surround the usage of a tool. The older I've gotten as a person who was deeply tribalistic about Ruby on Rails when I got into it and Ember when I got into it, because I love tribes, I think tribes are awesome and it's a way to make friends but when you really lean into that, the costs are too high and experimenting with other technologies and noticing flaws in your own technology is not only not a betrayal, it's actually critical to your growth as a developer. The more people that do that, like Chris was saying, the better both of those ecosystems will get.

CHARLES: Absolutely because having spent as much time in React as I have, I really appreciate the precious things about Ember. It will make you appreciate the things that you hold dear. It will make you appreciate the really, really, really special things about the tool that you're using and at the same time, it will highlight the weaknesses which you can immediately use to feedback and make your tool better. It really is a win-win situation.

TARAS: I just want to do a little plug before we close up. I think the feels of working with Ember is actually gone into microstates and we're still getting our things together to make microstates look accessible and usable by everyone but that feeling of pleasure that you get from working with Ember and just things just being there for you, like we really want to reproduce that and make that available in React community and the stuff that we do in microstates is actually really designed for that.

CHRIS: Yeah, I see that in BigTest too as well. That's definitely another place where it's like, "These people definitely used to spend time in Ember and they're now in React-land." It's cool to see that stuff getting ported over.

CHARLES: Absolutely because it fundamentally changes your taste. Working with an application that doesn't have like a bolted on testing framework is like eating water soup. You just can't enjoy your life. It really is flavored everything that we do. On that note, we can go ahead and wrap up. There actually is some pretty exciting news. We're actually going to be launching a BigTest launcher. Up until this point, you kind of had to roll your own using BigTest for your assertions but using something like Karma to actually launch the browsers and we're actually launching our own launcher. I guess we've written our own launcher and we're going to be pushing it to NDM, not to overload the word launch.

You can look for that in the next couple of weeks. There's going to be a CLI that ships with BigTest to help you do even more set up, to make it so that you can just drop BigTest right into your application, whether it's jQuery, React, Ember, you name it. That should be really, really fun. Be looking for that and with that, if anybody has any other remarks...

BRANDON: If people are coming through RubyConf this year, I'll be there talking about management stuff. That's my only near-future conference stuff coming up. Hope to see some of the more Ruby-flavored folks out there.

CHARLES: All right-y. Definitely, go to every single talk that Brandon ever gives. You won't regret it. I can base that on very dear personal experience. You won't be disappointed. You know, not to put the pressure on or anything like that but you could never put any more pressure on Brandon than he puts on himself. With that, we will say good bye. Bye Chris, bye Brandon. Thank you so much. This is a great conversation. It certainly clarified a lot in my mind --

TARAS: Yes, same here.

CHARLES: -- About these problems. With that, we will say goodbye. Thank you for listening The Frontside Podcast. Please get in touch with us at @TheFrontside on Twitter or contact at Frontside.io on email. We do a range of custom services from full stack project development to JavaScript mentoring, to as you go JavaScript help desks kind of stuff. If you need to reach out to an expert, please get in touch. Our podcast as always is produced by the inimitable, Mandy Moore. Thank you very much and we'll see you all next time.

Listen to our podcast:

Listen on Apple Podcasts