We love pushing the boundaries on the web, and so we’ve decided to try something new. You probably have heard of JAMstack — the new web stack based on JavaScript, APIs, and Markup — but what does it mean for your workflow and when does it make sense in your projects?

As a part of our Smashing Membership, we run Smashing TV, a series of live webinars, every week. No fluff — just practical, actionable webinars with a live Q&A, run by well-respected practitioners from the industry. Indeed, the Smashing TV schedule looks pretty dense already, and it’s free for Smashing Members, along with recordings — obviously.

We’ve kindly asked Phil Hawksworth to run a webinar explaining what JAMStack actually means and when it makes sense, as well as how it affects tooling and front-end architecture. The one hour-long webinar is now available as well. We couldn’t be happier to welcome Phil to co-MC our upcoming SmashingConf Toronto (June 25-26) and run JAMStack_conf London, which we co-organize on July 9-10 this year as well. So, let’s get into it!

Phil Hawksworth: Excellent, okay, well let’s get into it then. Just by way of a very quick hello, I mean I’ve said hello already, Scott’s given me a nice introduction. But yes, I currently work at Netlify, I work in the developer experience team there. We are hopefully going to have plenty of time for Q&A but, as Scott mentioned, if you don’t get a chance to ask questions there, or if you just rather, you can ping them directly at me on Twitter, so my Twitter handle is my names, it’s Phil Hawksworth, so any time you can certainly ask me questions about JAMstack or indeed anything on Twitter.

Phil Hawksworth: But I want to start today just by kind of going back in time a little bit to this quote which really resonates very, very strongly with me. This is a quote from the wonderful Aaron Schwartz who, of course, contributed so much to the Creative Commons and the open web and he wrote this on his blog way back in 2002, and he said, “I care about not having to maintain cranky AOL server, Postgres and Oracle installs.” AOL server, I had to look up to remind myself was an open source web server at the time.

Phil Hawksworth: But this chimes really strongly with me. I also don’t want to be maintaining infrastructure to keep a blog alive, and that’s what he was talking about. And it was in this blog post on his own blog and it was titled “Bake, Don’t Fry”. He was picking on a term that someone who’d built a CMS recently had started to use, and he kind of popularized this term about baking (Bake, Don’t Fry); what he’s talking about there is pre-rendering rather than rendering on demand, so baking the content ahead of time, rather than frying it on demand when people come and ask for it — getting things away from request time and into kind of build time.

Phil Hawksworth: And when we’re talking about pre-rendering and rendering, what we mean by that is we’re talking about generating markup. I feel a bit self-conscious sometimes talking about kind of server render or isomorphic rendering or lots of these kind of buzzwordy terms; I got called out a few years ago at a conference, Frontiers Conference in Amsterdam when I was talking about rendering on the server and someone said to me, “You mean generating HTML? Just something that outputs HTML?” And that’s, of course, what I mean.

Phil Hawksworth: But all of this kind of goes a long way towards simplifying the stack. When we think about the stack that we serve websites from; I’m all about trying to simplify things, I’m super keen on trying to simplify the stack. And that’s kind of at heart of this thing called “JAMstack” and I want to try and explain this a little bit. The “JAM” in JAMstack stands for JavaScript, APIs and Markup. But that’s not enough really to help us understand what it means — what on earth does that really mean?

Phil Hawksworth: Well, what I want to try and do in the next half hour or so, is I want to kind of expand on that definition and give more of a description of what JAMstack is. I want to talk a bit about the impact and the implications of JAMstack, and, you know, think about what that can give us as to why we might choose it. Along the way, I’m going to try to mention some of the tools and services that will be useful, and hopefully, I’ll wrap up with some resources that you might want to dig into and perhaps mention some first steps to get you under way.

Phil Hawksworth: So, that’s the plan for the next half-hour. But, I want to, kind of, come back to this notion about simplifying the stack, because, hopefully, people who join this or have come to watch this video later on, maybe you’ve got a notion of what JAMstack is, or maybe it’s a completely new term, and you’re just curious. But, ultimately, there are a lot of stacks out there, already. There are lots of ways that you can deliver a website. It feels like we’ve been building different types of infrastructure for a really long time, whether that’s a LAMP stack or the MAMP stack, or the — I don’t know — the MEAN stack. There’s a bunch of them floating by on the screen here. There are lots and lots of them.

Phil Hawksworth: So, why on earth would we need another one? Well, JAMstack is, as I mentioned, is JavaScript/API/Markup, but if we try and be a tiny bit more descriptive, JAMstack is intended to be a modern architecture, to help create fast and secure sites and dynamic apps with JavaScript/APIs and pre-rendered markup, served without web servers, and it’s this last point which is, kind of, something that sets it apart and maybe, makes it a little bit more, kind of, interesting and unusual.

Phil Hawksworth: This notion of serving something without web servers, that sounds either magical or ridiculous, and hopefully, we’ll figure out what along the way. But to try and shed some light over this and describe it in a little bit more detail, it’s sometimes useful to compare it to what we might think of as a traditional stack, or a traditional way of serving things on the web. So, let’s do that just for a second. Let’s just walkthrough, perhaps, what a request might look like as it gets serviced in a traditional stack.

Phil Hawksworth: So, in this scenario, we got someone opening up a web browser and making a request to see a page. And maybe that request hits a CDN, but probably, more likely, it hit some other infrastructure that we are hosting — as the people who own this site. Maybe we tried to make sure that this is going to scale under lots of load because we, obviously, want a very popular and successful sight. So, perhaps we got a load balancer, that has some logic in it, which will service that request to one of a number of web servers that we’ve provisioned and configured and deployed to. There might be a number of those servers.

Phil Hawksworth: And those servers will execute some logic to say, “Okay, here’s our template, we need to populate that with some data.” We might get our data from one of a number of database servers that will perform some logic to look up some data, return that to the web server, create a view that we then pass back through the load balancer. Perhaps, along the way, calling off to CDN, stashing some assets in the CDN, and I should clarify, a CDN is a Content Delivery Network, so a network of machines distributed around the Internet to try and get request service as close to possible to the user and add things, like caching.

Phil Hawksworth: So, we might stash some assets there, and ultimately, return a view into the browser, into the eyeballs of the user, who gets to then experience the site that we built. So, obviously, that’s, either, an oversimplification or a very general view of how we might service a request in a traditional stack. If we compare that to the JAMstack, which is servicing things in a slightly different way, this is how it might look.

Phil Hawksworth: So, again, same scenario, we’re starting in a web browser. We’re making a request for a view of the page, and that page is already in a CDN. It serves statically from there, so it’s returned to the user, into the browser, and we’re done. So, obviously, a very simplified view, but straight away, you can start to see the differences here in terms of complexity. In terms of places that we need to manage code, deeply code, all of those different things. So, for me, one of the core attributes one of a JAMstack, is that it means that you’re building a site that’s capable of being served directly from a CDN, or even from a static file server. CDN is something that we might want to put in place to handle load, but ultimately, this could be served directly from any kind of static file server, kind of static hosting infrastructure.

Phil Hawksworth: JAMstack, kind of, offers an opportunity to reduce complexity. Just comparing those two parts of the diagram that we’ll come back to a few times, over the course of the next half hour, you can see that it’s an opportunity to reduce complexity and reduce risk. And so, it means that we can start to enjoy some of the benefits of serving static assets, and I’m going to talk about what those are a little bit later on. But you might be looking at this and thinking, “Well, great, but isn’t this just the new name for static websites?” That’s a reasonable thing to level at me when I’m saying, “We’re going to serve things statically.”

Phil Hawksworth: But I want to come back to that. I want to talk about that a little bit more, but first of all, I want to, kind of, talk about this notion of stacks and what on earth is a stack, anyway? And I think of a stack as the layers of technology, which deliver your website or application. And we’re not talking about the build pipeline, or the development process, but certainly the way we serve sites can have a big impact on how we develop and how we deploy things, and so on.

Phil Hawksworth: But here, we’re talking about the technology stack, the layers of technology, that actually deliver your website and your application to the users. So, let’s do another little comparison. Let’s talk about the LAMP stack for a second.

Phil Hawksworth: The LAMP stack, you may remember, is made up of an apache web server, for doing things like the HCP routing and the serving of static assets. PHP, for some pre-processing, so pretty hyper-text processing; applying the logic, maybe building the views for the templates and what have you. And has some access to your data, by my NISQL, and then LINUX is the operating system that sits beneath all of that, keeps that all breathing. We can wrap that up together notionally as this web server. And we may have many of these servers, kind of, sitting together to serve a website.

Phil Hawksworth: That’s a, kind of, traditional look at the LAMP stack, and if we compare that to the JAMstack, well, here, there’s a critical change. Here, we’re actually moving up level, rather than thinking about the operating system and thinking about how we run the logic to deliver a website. Here we’re making an assumption that we’re going to be serving these things statically. So, we’re doing the ACP routing, and the serving of assets from a static server. That can be reasonably done. We got very good at this, over the years, building ways to deliver static websites, or static assets.

Phil Hawksworth: This might be a CDN, and again, we’ll talk about that in a moment. But the area of interest for us, is happening more in the browser. So, here, this is where our markup is delivered and is parsed. This is where JavaScript can execute, and this is happening in the browser. In many ways, the browser has become the runtime for the modern web. Rather than having the runtime in the server infrastructure, itself, now we’ve moved that up a level, closer to the user, and into the browser.

Phil Hawksworth: When it comes to accessing data, well, that’s happening through, possibly, external APIs, making calls via JavaScripts to these external APIs to get server access, or we can think APIs as the browser APIs, being able to interact with JavaScript with capabilities right there in your browser.

Phil Hawksworth: Either way, the key here about the JAMstack is that, we’re talking about things that are pre-rendered: they’re served statically and then, they maybe progressively enhanced in the browser to make use of browser APIs, JavaScripts, and what have you.

Phil Hawksworth: So, let’s just do this little side-by-side comparison here. Again, I just want to kind of reiterate that the JAMstack has moved up a level to the browser. And if we see the two sides of this diagram, with the LAMP stack on the left and effectively, the JAMstack on the right, you might even think that, well, even when we were building things with the LAMP stack, we’re still outputting mark-up. We’re still outputting JavaScript. We might still be accessing APIs. So, in many ways, the JAMstack is almost like a subset of what we were building before.

Phil Hawksworth: I used to sometimes talk about JAMstack as the assured stack, because it’s assures a set of tools and technologies that we need to deliver a site. But, either way, it’s a much simplified way of delivering a site that, kind of, does away with the need for things to execute and perform logic at the server at request time.

Phil Hawksworth: So, this can do a lot of things. This can, kind of, simplify deployments and again, I’m going to call back to this diagram from time-to-time. If we think about how we deploy our code and our site, for every deploy, from the very first one, through the whole development lifecycle, all the way through the life of the website. On the traditional stack, we might be having to change the logic and the content for every box on that diagram.

Phil Hawksworth: Whereas, in the JAMstack, when we’re talking about deploying, we’re talking at getting things to the CDN, getting things to the static server, and that’s what the deployment entails. The build, the kind of logic that runs the build — that can run anywhere. That doesn’t need to run in the same environment that’s hosting the web server. In fact, it doesn’t! It starts the key to the JAMstack. We put the separation at what happens at request time, serving these static assets, versus what happens at build time, which can be your logic that you run to build and then to the deployment.

Phil Hawksworth: So, this kind of decoupling is a key thing, and I’m going to come back to that later on. We’ve got very good at serving static files, and getting things to a CDN or getting things to the file system (the file server) is somewhere that we’ve seen huge, kind of, advancement over the last few years. There are a lot of tools and processes, now, that can help us do this really well. Just to call out a few services that can serve static assets well and give workflows to getting your build to that environment, they’re the usual suspects that you might imagine the big clouds infrastructure providers, like Azure, AWS, and Google Cloud.

Phil Hawksworth: But then, there are others. So, the top one on the right is a service called Surge, which has been around for a few years. It allows you to run a command in your build environment and deploy your assets through to their hosting environment. Netlify, the next one down, is where I work and we do very much the same thing but we have different automation as well. I could go into it another time. And the one on the bottom, another static hosting environment site, called Now.

Phil Hawksworth: So, there’s a bunch of different options for doing this, and all of these spaces provide different tooling for getting to the CDN, as quickly as possible. Getting your sites deployed in the most seamless way that we can. And they all have something in common where they’re building on the principal of running something locally. I often think of a static site generator as something that we might run in a build which when we run that, it takes things like content and templates and maybe, data from different services and it outputs something which can be served statically.

Phil Hawksworth: We can preview that locally in our static server. Something that is kind of simple to do on any local development environment, and then the process of deploying that is getting that to the hosting environment and ideally, out to a CDN in order to get, kind of, scale. So, with that kind of foundation laid out, I want to address a bit of a common misconception when it comes to JAMstack sites. And I didn’t do myself any favors by opening this up as describing JAMstack sites as being JavaScript, APIs, and Markup. Because the common misconception is that every JAMstack site has to be JavaScript and APIs, and Markup, but this kind of thing that we’ve overlooked is that we don’t have to use all three — every one one of these is, kind of, optional. We can use as much, or as little of these as we like. In the same way that a LAMP stack site wouldn’t necessarily need to be hitting a data base. Now, I’ve built things in the past that are served by an apache server, on a Linux machine, and I’ve been using PHP, but I haven’t been hitting a database and I wouldn’t start to rename a stack necessarily for that.

Phil Hawksworth: So, if we think about what is a JAMstack site, then it could be a bunch of different things. It might be something that’s built out with a static site generator, like Jekyll, pulling content from YAML files to build a site that has no JavaScript, doesn’t hit APIs at all, and we serve it on something, like GitHub Pages. Or, that would be a JAMstack site. Or maybe we’re using a static site generator, like Gatsby, which is, rather in a Ruby environment for Jekyll, now this is a JavaScript static site generator built in the React ecosystem.

Phil Hawksworth: That might be pulling content again, and it’s organizing Markdown files. It might be enriching that with calls to APIs, GraphQL’s APIs. It might be doing things in the browser, like doing JavaScript hydration of populating templates right there in the browser. And it might be served on Amazon S3. Is that a JAMStack site? Yeah, absolutely!

Phil Hawksworth: Moving on to a different static site generator, Hugo, which is built with Go! Again, we might be organizing content in Markdown files, adding interactions in the browser using JavaScript. Maybe not calling any external APIs and maybe hosting that on Google Cloud. Is it JAMstack? Absolutely! You see, I’m building to a theme here. Using something like Nuxt, another static site generator, now built in the View ecosystem. Maybe that’s pulling content from different adjacent files? Again, we might be using JavaScript interactions in the browser, perhaps calling APIs to do things like e-Commerce, serving it another static site. Another hosting infrastructure like Netlify, is it a JAMstack? Yes, it is!

Phil Hawksworth: But we might even go, you know, go off to this side end of the scale, as well. And think about a handmade, progressive web app that we’ve built artisanally, hand-rolled, JavaScript that we built ourselves. We’re packaging it up with webpack. We’re maybe using JavaScript web tokens and calling out to APIs to do authentication, interacting with different browser APIs, hosting it on Azure. Yes, that’s JAMstack as well!

Phil Hawksworth: And, you know, all of these things, and many more, can be considered JAMstack, because they all share one attribute in common and that is none of them are served with an origin server. None of them have to hit a server that performs logic at request time. These are being served as static assets, and then enriched with JavaScript and calls to APIs, afterwards.

Phil Hawksworth: So, again, I just want to reiterate that a JAMstack means it’s capable of being served directly from the CDN. So, I want to just call out some of the impacts and implications of this, because why would we want to do this? Well, the first notion is about security, and we’ve got a greatly reduced surface area for attack, here. If we think about (coming back to this old diagram again), the places where we might have to deal with an attack, we have to secure things like the load balancer, the webservers, the database servers. All of these things, we have to make sure aren’t able to be penetrated by any kind of an attack and, indeed, the CDN.

Phil Hawksworth: If the more pieces we can take out of this puzzle, the fewer places that can be attacked and the fewer places we have to secure. Having few moving parts to attack is really very valuable. At Netlify, we operate our own CDNs, so we get the luxury of being able to monitor the traffic that comes across it, and even though all of the sites hosted on Netlify, all of the JAMstack sites that you might imagine, none of them have a WordPress admin page on them because it’s completely decoupled. There is no WordPress admin, but we see a huge volume of traffic, probing for things like WP Admin, looking for ways in, looking for attack vectors.

Phil Hawksworth: I really love some of the things that Kent C. Dodds has done. I don’t know if you are familiar with the React community, you’ve probably encountered Kent C. Dodds in the past. He doesn’t use WordPress, but he still routes this URL, the WPAdmin. I think he used to route it through to a Rick Roll video on YouTube. He’s certainly been trolling people who have gone probing for it. But, the point is, by decoupling things in that way and, kind of, moving things that happen, build time from what happens in request time, we can just drastically reduce our exposure. We’ve got no moving parts at request time. The moving parts are all completely decoupled at build time. Potentially, on completely, well, necessarily on completely different infrastructure.

Phil Hawksworth: This, of course, also has an impact on performance, as well. Going back to our old friend here, the places we might want to try and improve performance across the stack here, when there’s logic that needs to be executed at these different places. The way that this often happens in traditional stacks is, they start to add layers, add static layers, in order to improve performance. So, in other words, try and find ways that we can start to behave as if it’s static. Not have to perform that logic at each level of the stack in order to speed things up. So, we’re starting to introduce things like caching all over the infrastructure and obvious places we might find to do that is in the web server, rather than perform that logic, we want to serve something immediately without performing that logic.

Phil Hawksworth: So, it’s kind of like a step towards, kind of, being pseudo-static. Likewise in database servers, we want to add caching layers to cache-com queries. Even in the low balance, the whole CDN, you can think of as a cache. But on the traditional stack, we need to figure out how to manage that cache, because not everything will be cached. So, there’s going to some logic about. What needs to be dynamically populated versus what can be cached. And the JAMstack model, everything is cached. Everything is cached from the point that the deployment is done, so we can think about it completely differently.

Phil Hawksworth: This, then, starts to, kind of, hint through to scaling, as well. And by scale, I’m talking about, how do we handle large loads of traffic? Traditional stacks will start to add infrastructure in order to scale. So, yes, to caching. We’re starting to put in place in our traditional stack. That will help — to a degree. What typically happens is, in order to handle large loads of traffic, we’ll start expanding the infrastructure and starting to add more servers, more pieces to handle these requests, costing these things out and estimating the load is a big overhead and it can be a headache for anyone doing technical architecture. It certainly was for me, which is why I was starting to lean much more towards doing the JAMstack approach where I just know that everything is served from the CDN, which is designed by default to handle scale, to handle performance right out of the gate.

Phil Hawksworth: So, I also want to give a nod to developer experience, and the impact this can have there. Now, developer experience should never be seen as something which trumps user experience, but I believe that a good developer experience can reduce friction, can allow for developers to do a much better job of building up to great user experiences!

Phil Hawksworth: So, when we think about where the developer experience lives, and where the areas of concern for a developer are here: well, in a traditional stack, we might need to think about how we get the code to all of these different parts of the infrastructure, and how they all play together. In the JAMstack world, really, what we’re talking about is this box here at the bottom. You know, how do we ran the build and them, how do we automate a deployment to get something served in the first place? And the nice thing is, that in the JAMstack model, what you do in that build is completely up to you.

Phil Hawksworth: That’s a really well-defined problem space, because ultimately, you’re trying to build something you can serve directly from a CDN. You want to pre-render something, using whatever tools you like: whether it’s a static site generator built in Ruby or Python or JavaScript or Go or PHP, you have the freedom to make that choice. And so, that can create a much nicer environment for you to work in. And also, it creates an opportunity to have real developer confidence because a real attribute of JAMstack sites is that they can be much more easily served as immutable and atomic deployment.

Phil Hawksworth: And I want to, kind of, jump away from the slides just for a moment, to describe what that means, because an immutable deployment and an atomic deployment can… (that can just sound a little bit like marketing speak) but what I’m going to do, is I’m going to jump into my browser. Now … actually, I’m going to go back for a second. Let me… just do this.

Phil Hawksworth: Here we are. This will be easier. Jumping right into the page. Now, Scott, you will tell me, Scott, if you can’t see my browser, I’m hoping. Now, assuming everyone can see my browser.

Scott: Everything looks good.

Phil Hawksworth: Excellent! Thank you very much!

Phil Hawksworth: So, what I’m doing here, is I’m using Netlify as an example, as an example of the service. But, this is an attribute which is common to sites that can be hosted, statically. So, when we talk about an immutable deployment, what we mean is, that rather each deployment of code having to touch lots of different parts of the infrastructure, and change lots of things, here we’re not mutating the state of the site on the server. We’re creating an entirely new instance of the site for every deployment that’s happened. And we can do that because the site is a collection of static assets.

Phil Hawksworth: Here, I’m looking at the deployment that have happened from my own website. I’ll give you a treat. There you are, that’s what it looks like. It’s just a blog, it doesn’t look like anything particularly remarkable or exciting. It’s a statically generated blog, but what I have here is every deployment that’s ever happened, lives on in perpetuity, because it’s a collection of static assets that are served from a CDN. Now, I could go back as far as my history can carry me and go and look at the site, as it was back in… when was this? This was August, 2016. And by virtue of it being a set of static assets, we’re able to host this on its own URL that lives on in perpetuity and if I even wanted to, I could decide to go in and publish that deployment.

Phil Hawksworth: So, now, anyone’s who’s looking at my website, if I go back to my website here, if I refresh that page, now that’s being served directly from the CDN with the assets that were there before. Now, I can navigate around again. Here, you can see. Look, I was banging on about this, I was using these terrible terms like isomorphic rendering and talking about the JAMstack back in 2016. So, this is now what’s being served live on my site. Again, because there are mutual deployments that just live on forever. I’m going to just put my own, kind of, peace of mind, I’m going to — is this the first page? Yeah. I’m going to go back to my latest deployment. I’ll have to shut again, and get me back into the real world. Let me just make sure this is okay.

Phil Hawksworth: Okay! Great! So, then now, this is back to serving my previous version, or my latest version of the site. I’ll hop back to keynote. So, this is possible because things are immutable and atomic. The atomic part of that means, again, that the deployment is completely contained. So, you never get the point where some of the assets are available on the web server, but some of them won’t. Only when everything is there in context and everything is there, complete, do we toggle the serving of the site to the new version. Again, this is the kind of thing you can do much more easily if you’re building things out as a JAMstack site that serves directly from the CDN as a bunch of assets.

Phil Hawksworth: I noticed that my timer has reset, after going back and forward from keynote, so I think I have about six or seven minutes left. Tell me, Scott, if—

Scott: So, yeah, we’re still good for about ten minutes.

Phil Hawksworth: Ten minutes? Okay, wonderful!

Scott: There’s no rush.

Phil Hawksworth: Thank you, that should be good!

Phil Hawksworth: So, just switching gear a tiny bit and talking about APIs and services (since APIs is part of JAMstack), the kind of services that we then might be able to use is mostly JAMstack. You know, we might be using services that we built in-house, or we might be using bought-services. There are lots of different providers who can do things for us, and that’s because that’s their expertise. Through APIs, we might be pulling in content from content management systems as a service, and there’s a bunch of different providers for this, who specialize in giving a great content management experience and then, exposing the content through API, so you used to be able to pull them in.

Phil Hawksworth: Likewise, there are different ways that you can serve assets. People like Cloudary are great at this, for doing image optimization, serving assets directly to your sites, again, through APIs. Or what about e-Commerce? You know, there are places like Stripe or Snipcart that can provide us API, so that we don’t have to build these services ourselves and get into the very complex issues that come with trying to build an e-Commerce engine. Likewise, identity, from people like Auth0 who are using Oauth. There are lots of services that are available and we can consume these things through APIs, either in the browser or at build time, or sometimes, a combination of both.

Phil Hawksworth: Now, some people might think, “Well, that’s fine, but I don’t want to give the keys to the kingdom away. I don’t want to risk giving these services over to external providers,” and to that, I say, well, you know, vendors who provide a single service really depend on its success. If there’s a company that’s core business, or entire business, is in building-out an e-Commerce engine, an e-Commerce service for you, they’re doing that for all of their clients, all of their customers, so they really depend on its success and they have the specialist skills to do that.

Phil Hawksworth: So, that kind of de-risks it from you a great deal. Especially when you start to factor in the fact that you can have your technical and service-level contracts to give you that extra security. But it’s not all about bought services, it’s also about services you can build yourselves. Now, there’s a number of ways that this can happen, but sometimes, you absolutely need a little bit of logic on the server. And so far, I’ve just been talking about taking the server out of the equation. So, how do we do that?

Phil Hawksworth: Well, this is where serverless can really come to the rescue. Serverless and JAMstack, they just fit together really beautifully. And when I’m talking about serverless, I’m talking about no functions as a service. I know that serverless can be a bit of a loaded term, but here, I’m talking about functions as a service. Because functions as a service can start to enable a real micro-services architecture. Using things like AWS Lambda or Google Cloud functions or similar functions as a service, can allow you to build out server infrastructure without a server. Now, you can start deploying JavaScript logic to something that just runs on demand.

Phil Hawksworth: And that means, you can start supplementing some of these other services with, maybe, very targeted small services you build yourselves that can run the serverless functions. These kind of smaller services are easier to reason about, understand, build out and they create much greater agility. I want to just mention a few examples and results from JAMstack sites. I’m not going to go down the server list avenue too much, right now. We can, maybe, come back to that in the questions. I really just kind of want to switch gear and, thinking about time a little bit, talk about some examples and results.

Phil Hawksworth: Because there are some types of sites that lend themselves in a very obvious way to a JAMstack site. Things like the documentation for React, or Vuejs, those [inaudible 00:32:40], pre-rendered JAMstacks sites. As do sites for large companies, such as Citrix, this is a great example of Citrix multi-language documentation. You can actually view the video from the JAMstack conference that happened in New York, where Beth Pollock had worked on this project, talked about the change that went on in that project. From building on traditional, non-enterprised infrastructure to a JAMstack approach and building with Jekyll, which is not necessarily the fastest generating static site generator, but still, they saw a huge improvement.

Phil Hawksworth: Beth talked about the turnaround time for updates went from weeks to minutes. Maybe people are kind of frowning at the idea of weeks for updates to sites, but sometimes in big complex infrastructure, with lots of stakeholders and lots of moving parts, this really is the situation we’re often finding ourselves in. Beth also talked about estimating the annual cost savings for this move to a JAMstack site. To get the site running properly, estimated their savings of around 65%. That’s huge kind of savings. Changing gear to a slightly different type of site, something a little closer to home, Smashing Magazine itself is a JAMstack site, which might be a little bit surprising, because on one hand, yes, there’s lots of articles and it’s also content, which is published regularly, but not every minute of the day, for sure.

Phil Hawksworth: So, that might lend itself, perhaps, for something that’s pre-generated, but of course, there’s also a membership area and an event section, and a job board, and e-Commerce, and all of these things. This is all possible on the JAMstack because not only are we pre-rendering, but we’re also enriching things with JavaScript and the front end to call out to APIs, which let some of these things happen. The project that I think I saw Vitaly arrive in the call, so that’s going to be good, we might be able to come back to this in a few minutes.

Phil Hawksworth: But the project that migrated, Smashing Magazine onto a JAMstack approach, I believe, simplified the number of platforms from five, effectively down to one. And I’m using Vitaly’s words directly here: Vitaly talked about having some caching issues, trying to make the site go quickly, using probably every single WordPress caching plug-in out there, and goodness knows, there are a few of them! So, Smashing Magazine saw an improvement in performance, time to first load went from 800 milliseconds to 80 milliseconds. Again, I’m simplifying the infrastructure that served the site up in the first place. So, it’s kind of interesting to see the performance gains that can happen there.

Phil Hawksworth: Another totally different type of site. This is from the Google Chrome team, who built this out to demonstrate at Google I/O this year. This very much feels like an application. This is Minesweeper in a browser. I apologize if you’re watching me play this. I’m not playing it while talking to you; I recorded this sometime ago and it’s agony to see how terrible I seem to be at Minesweeper while trying to record. That’s not a mine, that can’t be!

Phil Hawksworth: Anyway, we’ll move on.

Phil Hawksworth: The point of that is, this is something that feels very much more like an app, and it was built in a way to be very responsible about the way it used JavaScript. The payload to interactive was 25KB. This progressively would download and use other resources along the way, but it meant that the time to interact was under five seconds on a very slow 3G network. So, you can be very responsible with the way you use JavaScript and still package up JavaScript, as part of the experience for a JAMstack site.

Phil Hawksworth: So, I’m kind of mindful of time. We’re almost out of time, so what is the JAMstack? Well, it’s kind of where we started from. JAMstack sites are rendered in advance: they’re not dependent on an origin server (that’s kind of key), and they may be progressively enhanced in the browser with JavaScript. But as we’ve shown, you don’t have to use JavaScript at all. You might just be serving that statically, ready to go, without that. It’s an option available to you.

Phil Hawksworth: This key tenant, I think of, JAMstack sites is they’re served without web service. They’re pre-rendered and done in advance.

Phil Hawksworth: If you’re interested in more, it’s already been mentioned a little bit, there is a conference in London in July — July 9th and 10th. The speakers are going to be talking about all kinds of things to do with performance in the browser, things that you can do in the browser, results of building on the JAMstack and things to do with serverless.

Phil Hawksworth: There’s also a bunch of links in this deck that I will share, after this presentation, including various bits and pieces to do with static site generation, things like headless CMS, the jamstack.org site itself, and a great set of resources on a website called “The New Dynamic” which is just always full of latest information on JAMstack. We’re out of time, so I’ll wrap it up there, and then head back across to questions. So, thanks very much for joining and I’d love to take questions.

Scott: Thanks, Phil. That was amazing, thank you. You made me feel quite old when you pulled up the Minesweeper reference, so—

Phil Hawksworth: (laughs) Yeah, I can’t take any credit for that, but it’s kind of fascinating to see that as well.

Scott: So, I do think Vitaly is here.

Vitaly: Yes, always in the back.

Phil Hawksworth: I see Vitaly’s smiling face.

Vitaly: Hello everyone!

Phil Hawksworth: So, I’m going to hand it over to Vitaly for the Q&A, because I seem to have a bit of a lag on my end, so I don’t want to hold you guys up. Vitaly, I’ll hand it over to you.

Scott: Okay. Thanks, Scott.

Vitaly: Thanks, Scott.

Vitaly: Hello—

Vitaly: Oh, no, I’m back. Hello everyone. Now Scott is back but Phil is gone.

Scott: I’m still here! Still waiting for everything.

Vitaly: Phil is talking. Aw, Phil, I’ve been missing you! I haven’t seen you, for what, for days, now? It’s like, “How unbelievable!” Phil, I have questions!

Vitaly: So, yeah. It’s been interesting for us, actually, to move from WordPress to JAMstack — it was quite a journey. It was quite a project, and the remaining moving parts and all. So, it was actually quite an undertaking. So, I’m wondering, though, what would you say, like if we look at the state of things and if we look in the complexes, itself, that applications have. Especially if you might deal with a lot of legacy, imagine you have to deal with five platforms, maybe seven platforms here, and different things. Maybe, you have an old legacy project in Ruby, you have something lying on PHP, and it’s all kind of connected, but in a hacky way. Right? It might seem like an incredible effort to move to JAMstack. So, what would be the first step?

Scott: So … I mean, I think you’re absolutely right, first of all. Re-platforming any site is a big effort, for sure. Particularly if there’s lots of legacy. Now, the thing that I think is kind of interesting is an approach that I’ve seen getting more popular, is in identifying attributes of the site, parts of the site, that might lend themself more immediately to being pre-generated and served statically than others. You don’t necessarily have to everything as a big bang. You don’t have to do the entire experience in one go. So, one of the examples I shared, kind of, briefly was the Citrix documentations site.

Scott: They didn’t migrate all of Citrix.com across to being JAMstack. They identified a particular part that made sense to be pre-rendered, and they built that part out. And then, what they did was they started to put routing in front of all the requests that would come into their infrastructure. So, it would say, “Okay, well, if it’s in this part of the past the domain, either in the sub-domain or maybe, through to a path, route that through something which is static, then the rest of it, pass that through to the rest of the infrastructure.

Scott: And I’ve seen that happen, time and time again, where with some intelligent redirects, which, thankfully, is something that you can do reasonably simply on the JAMstack. You can start to put fairly expressive redirect engines in front of the site. It means that you can pass things through just that section of the site that you tried to take on as a JAMstack project. So, choosing something and focusing on that first, rather than trying to do a big bang, where you do all of the legacy and migration in one. I think that’s key, because, yeah, trying to do everything at once is pretty tough.

Vitaly: It’s interesting, because just, I think, two days, maybe even today, Chris Coyier wrote an article renaming JAMstack to SHAMstack where, essentially, it’s all about JavaScript in which we need think about steady coasting, because JavaScript could be hosted steadily as well. And it’s interesting, because he was saying exactly that. He actually — when we think about JAMstack, very often, we kind of tend to stay in camps. It’s either, fully rendered and it lives a static-thing. Somewhere there, in a box and it’s served from a city and that’s it, or it’s fully-expressive and reactive and everything you ever wanted. And actually, he was really right about a few things, like identifying some of the things that you can off-load, to aesthetic-side, generated assets, so to say, to that area.

Vitaly: And, JAMstackify, if you might, say some of the fragments of your architecture. Well, that’s a new term, I’m just going to coin, right there! JAMstackify.

Phil Hawksworth: I’m going to register the domain quickly, before anybody else.

Phil Hawksworth: And it’s a nice approach. I think, it kind of makes my eye twitch a little bit when I hear that Chris Coyier has been redubbing it the SHAMstack, because it makes it sound like he thinks it’s a shambles. But I know that he’s really talking about static-hosting and markup, which I—

Vitaly: Yes, that’s right.

Phil Hawksworth: I really like, because the term JAMstack can be really misleading, because it’s trying to cover so many different things and the point I was trying to, I probably hammered it many times in that slide, is that it can be all kinds of things. It’s so broad, but the key is pre-rendering and hosting the core of the sites statically. It’s very easy for us to get into religious wars about where it needs to be a React site. It has to be a React app, in order to be a JAMstack site, or it’s a React app, so it can’t be JAMstack. But, really, the crux of it is, whether you use JavaScript or not, whether you’re calling APIs or not, if you pre-render and get things into a static host that can be very performant, that’s the core of JAMstack.

Vitaly: Yes, absolutely.

Phil Hawksworth: We’re very fortunate that browser’s are getting so much more capable, and the APIs that are there within browser’s themselves can allow us to do much more as well. So, that kind of opens the doors even further, but it doesn’t mean that everything that we build as a JAMstack site has to make use of everything. Depending on what we’re trying to deliver, that’s how we should start to choose the tools that we’re playing with to deploy those things.

Vitaly: Absolutely. We have Doran here. Doran, I think I know, Doran. I have a feeling that I know Doran. He’s asking, “Do you expect serverless to be gravitating towards seamless integration with JAMstack from [inaudible 00:44:36]? What is referred to as the A in JAM.

Phil Hawksworth: That’s a great question, because I think, serverless functions are — they just go so well with JAMstack sites, because in many ways, in fact, I think someone once asked me if JAMstack sites are serverless, and so I squirmed about that question, because serverless is such a loaded term. But, in many ways, it’s bang-on because I was talking, time and time again, about there’s no origin server. There’s no server infrastructure for you to manage. In fact, I once wrote a blog post called “Web Serverless,” because the world needs another buzz term, doesn’t it?

Phil Hawksworth: And really, the kind of point of that was, yes, we’re building things without servers. We don’t want to have to manage these servers, and serverless functions, or functions as a service, just fits into that perfectly. So, in the instances that you do need an API that you want to make a request to, where really it’s not sensible for you to make that request directly from the browser. So, for instance, if you’ve got secrets, or keys, in that request, you might not want those requests — that information — ever exposed in the client. But we can certainly proxy those things, and typically, traditionally, what we need to do then, is spin-up a server, have some infrastructure that was effectively doing little more than handling requests, adding security authentication to it and passing those requests on, proxying them back.

Phil Hawksworth: Serverless functions are perfect for that. They’re absolutely ideal for that. So, I sometimes think of serverless functions, or functions of a service, almost as like an escape hatch, where you just need some logic on a server, but you don’t want to have to create an entire infrastructure. And you can do more and more with that, and stipe the development pipelines for, development workflows, for functions as a service is maturing. It’s getting more accessible for JavaScript developers to be able to build these things out. So, yeah, I really think those two things go together very nicely.

Vitaly: All right, that’s a very comprehensive answer. Actually, I attended a talk just recently, where a front-end engineer from Amazon was speaking about serverless and Lamda functions they’re using — I was almost gone. He was always speaking about Docker, and Kubernetes, and all those things, Devox World, I was sitting there, thinking, “How did he end up there. I don’t understand what’s going on!” I had no idea what’s going on.

Phil Hawksworth: Exactly, but the thing is, it used to be the… I was… I accepted that I didn’t understand any of that world, but I didn’t have any desire to, since that was for an entirely different discipline. And that discipline is still really important. You know, people who are designing infrastructure — that’s still really key. But, it just feels, now, that I’m tempted. As someone with a front-end development background, as a JavaScript developer, I’m much more tempted to want to play in that world, because the tools are coming, kind of, closer to me.

Phil Hawksworth: It’s much more likely that I might be able to use some of these things, and deliver things kind of safely, rather than just as an experiment of my own, which is where I used to be dappling. So, it feels like we’re becoming more powerful as web developers, which is exciting to me.

Vitaly: Like Power Rangers, huh?

Vitaly: One thing I do want to ask you, though, and this is actually something we discussed already, maybe, a week ago, but I still wanted to bring it up, because the one thing that you mentioned in the session was the notion of having a stand-alone instance of every single deploy, which is really cool. The question, though, is if you have a large assignment, with tens of thousands of pages, you really don’t want to redeploy every thing, every single time. So, essentially, if you have, like, if you’re mostly using the static side of things. So, we had this idea for a while and I know this is actually something that you brought up last time. The idea of atomic deployments.

Vitaly: Where you actually, literally, were served some sort of div between two different versions of snapshots of the set-up. So, if you say, change the header everywhere, then, of course, every single page has to be redeployed. But if you change, maybe, a component, like let’s say, carousel, that maybe effects only a 1000 pages, then it would make sense to redeploy 15000 pages. But only this 1000. So, can we get there? Is it a magical idea that’s out there, or is it something that’s quite tangible, at this point?

Phil Hawksworth: I think that is, probably, the Holy Grail for static site generators and this kind of model because, certainly, you’ve identified probably the biggest hurdle to overcome. Or the biggest ceiling that you bump into. And that is websites that have many, tens of thousands or hundreds of thousands, or millions of URLs — the notion that the build can become very long. Being able to detect which URL’s will change, based on a code change, is a challenge. It’s not insurmountable, but it’s a big challenge. Understanding what the dependency graph is across your entire site and then, intelligently deploying that — that’s really tough to do.

Phil Hawksworth: Because as you mentioned, a component change might have very far-reaching implications but you — it’s difficult, always, to know how that’s going to work. So, there are a number of static site generators, the projects that are putting some weight behind that challenge, and trying to figure out how they do partial-regeneration and incremental builds. I’m very excited that the prospect that that might get solved day, but at the moment, it’s definitely a big challenge. You can start to do things like try to logically sharred your site, and think about, again, kind of similar to the migration issue. Well, this section I know is independent in terms of its, kind of, some of the assets that it uses, or the type of content that lives there, so I can deploy them individually.

Phil Hawksworth: But that’s not ideal to me. That’s not really the perfect scenario. One of the approaches that I’ve explored a little bit, just as a proof of concept, is thinking about how you do things, like, making intelligent use of 404s. So, for instance, a big use case for very large signs, maybe news sites is, when they need a URL when a breaking news story happens, they need to be first to get it deploy out there. They need to get a URL up there. Things like the BBC News, you’ll see that the news story will arrive on the website, and then overtime, they’ll add to it, incrementally, but getting there first is key. So, having a build that takes 10 minutes, 20 minutes, whatever it’s going to be, that could be a problem.

Phil Hawksworth: But, if they’re content is abstracted and maybe used to have been called from an API. I mentioned content management systems that are abstracted, like Contentful, or Sanity, or a bunch of those. Anything that has a content API changes to that content structure that will trigger a build and we’ll go through the run, but the other thing that could happen is that, well, if you publish your URL for that, and then publicize that URL, even if the build hasn’t run, if someone hicks that URL, if the first stop on its 404 is instead of saying, “We haven’t got it,” is actually to hit that API directly, then, you can say, “Well, the build hasn’t finished to populate that yet, but I can do it in the client.” I can go directly to the API, get that, populate it in the client.

Vitaly: Hmm, interesting.

Phil Hawksworth: So, even while the build is still happening, you can start populating those things. And then, once the build’s finished, of course it wouldn’t hit a 404. You would hit that statically running page. So, there are techniques and there are strategies to tackle it, but still, it’s a very long, rambling answer, I’m sorry, but my conclusion is, yeah, that’s a challenge. Fingers crossed we’ll get better strategies.

Vitaly: Yeah, that’s great. So, I’m wondering, so, at this point, we really aren’t thinking, not just what the performance in terms of the content delivering, but those in performance in terms of the build speed. Like building deployment. So, this is also something that we’ve been looking into for, quite a bit of time now, as well.

Vitaly: One more thing I wanted to ask you. So, this is interesting, like this technique that you mentioned. How do you learn about this? This is just something people tend to publish on their own blogs or, is it some medium or is there a central repository where you can get some sort of case studios, of how JAMstack—how companies moved while unloading, or have failed to move to JAMstack.

Phil Hawksworth: So, it’s kind of maturing this landscape a little bit, at the moment. I mean, some of these examples, I think, I’m in a very fortunate position, I work somewhere that I’m in a role that I’m playing with the toys, coming up with interesting ways to use it and start experimenting with them. So, these proofs of concepts are, kind of, things that I get to experiment with and try to address these challenges. But the, I kind of mentioned earlier, a case study that was shown at the JAMstack conference in New York, and certainly, events like that, we’re starting to see best practices or industry practices and industry approaches being talked about at those kind of events. And certainly, I want to see more and work on more case studies to get in places like on Smashing Magazines, so that we can share this information much more readily.

Phil Hawksworth: I think, large companies and the enterprise space, is gradually adopting JAMstack, in different places, in different ways, but the world is still sloped to get out there, so I think, each time a company adopts it and shares their experience, we all get to learn from that. But I really want to see more and more of these case studies get published, so that we can lean particularly about how these kind of challenges are overcome.

Vitaly: Alright, so, then, maybe just the last question from me, because I always like to read questions. So, the JAMstack land, if you could change something, maybe there is something that you desperately would love to see, beyond deployments. Anything else that would be really making you very happy? That would make your day? What would that be? What’s on your wishlist, for JAMstack?

Phil Hawksworth: What a question. I mean, had we not talked about incremental builds, that would be—

Vitaly: We did. That’s too late, now. This card has been passed, already. We need something else.

Phil Hawksworth: So—

Vitaly: What I mean, like on a platform, if you looked at the back platform, there are so many exciting things happening. We have Houdini, we have web components coming, and everything, since you could be changing the entire landscape of all the right components. On the other side, we have all this magical, fancy world with SS NGS, and, of course, obviously, we also have single-page applications and all. What are you most excited about?

Phil Hawksworth: I’m going to be obtuse here, because there is so much stuff that’s going on, it’s exciting, and there is so many new capabilities that you can make use of in the browser. The thing that I really get excited about is people showing restraint (laughs) and as I said, dull answer, but I love seeing great executions that are done with restraint, in a thoughtful — about the wider audience. It’s really good fun, and gratifying to build with the shiniest new JavaScript library or the new browser API that does, I don’t know, scratch and sniff capabilities in the browser, which we desperately need, any day now.

Phil Hawksworth: But I really like seeing things that I know are going to work in many, many places. They’re going to be really performant, are going to be sympathetic to the browsers that exist — not just on the desks of CEOs and CTOs who got the snazzy toys, but also people who have got much lower-powered devices, or they’ve got challenging network conditions and those kinds of things. I like seeing interesting experiences, and rich experiences, delivered in a way that are sympathetic to the platform, and kind of, compassionate for the wider audience, because I think the web reaches much further than us, the developers, who build things for it. And I get excited by seeing interesting things done, in ways that reach more people.

Phil Hawksworth: That’s probably not the answer you were necessarily—

Vitaly: Oh, that’s a nice ending. Thank you so much. No, that’s perfect, that really is. All right, I felt everything went good! Thank you so much for being with us! I’m handing out to Scott!

Phil Hawksworth: Great!

Vitaly: I’m just here to play questions and answers. So, thank you so much, Phil! I’m still here, but Scott, the stage is yours, now! Maybe you can share with us what’s coming up next on Smashing TV?

Scott: I will, but first, Phil, I can’t wait to see how the implementation of scratch-and-sniff API work. Sounds very interesting. And, Vitaly, JAMstackify is already taken.

Vitaly: (dejected) Taken?! Can we buy it?

Scott: No, it exists!

Vitaly: Well, that’s too late. I’m always late.

Phil Hawksworth: That’s exciting in its own way.

Vitaly: That’s the story of my life. I’m always late.

Scott: Members coming up next, I believe, Thursday, the 13th, we have my ol’ pa, Zach Leatherman, talking about what he talks about best, which is fonts. So, he’s talking about the Five Y’s of Font Implementations. And then, I’m also very interested in the one we have coming up on the 19th, which is editing video, with JavaScript and CSS, with Eva Faria. So, stay tuned for both of those.

Scott: So, that is again, next Thursday, with Zach Leatherman, and then on the 19th, with Eva, who will be talking about editing video in JavaScript and CSS. So, on that note, Phil, I can’t see you anymore, are you still there?

Phil Hawksworth: I’m here!

Scott: On that note, thank you very much everyone! Also, is anybody in the, kind of, close to Toronto area? Or anybody that’s ever wanted to visit Toronto? We have a conference coming up at the end of June, and there’s still a few tickets left. So, maybe we’ll see some of you there.

Vitaly: Thank you so much, everyone else!

Vitaly: Oh, by the way, just one more thing! Maybe Phil mentioned it, but we also have the JAMstack Conference in London, in July. So, that’s something to watch out for, as well. But I’m signing off and going to get my salad! Not sure what you—

Scott: Okay, goodbye, everybody!

Vitaly: All right, bye-bye, everyone.

That’s A Wrap!

We kindly thank Smashing Members from the very bottom of our hearts for their continuous and kind support — and we can’t wait to host more webinars in the future.

Also, Phil will be MCing at SmashingConf Toronto 2019 next week and at JAMstack_conf — we’d love to see you there as well!

Please do let us know if you find this series of interviews useful, and whom you’d love us to interview, or what topics you’d like us to cover and we’ll get right to it.

(ra, il)