WP Engine and Atlas Headless CMS

WP Engine recently announced the launch of their Headless CMS hosting and frontend JavaScript stack, Atlas. This sounds really interesting so I took a look at their recent DE{CODE} 2021 conference videos on this topic to find out more.

As a little background we’ve used WP Engine at Studio 24 for a number of years, for enterprise WordPress hosting for some of our larger clients. We’ve always found their WordPress hosting solution rock solid, they have a bunch of very clever engineers developing solutions. When I attended a previous (in-person) WP Engine conference I was always surprised they hadn’t developed their own software to complement WordPress. Looks like that has now changed.

Atlas and headless

The keynote talk was a good summary by founder Jason Cohen on the reasons for investing in Headless. I liked the stat that 64% of enterprises are using headless. Many of us have been using decoupled architectures for years – just without calling it headless!

The key selling points are performance (with 10x page speeds than traditional WordPress), better security, ability to redesign without re-implementing your CMS, and developer freedom to work with any modern framework – as long it’s in JavaScript (for Atlas).

Performance and security I thoroughly agree with. Stripping back the frontend to only what’s required is clearly a great way to improve security and reduce attack vectors.

The ability to redesign a site using headless without having to reimplement your work in a CMS is also a big selling point, however, this assumes a solid frontend that does not also need to significantly change. I think this is the idea behind WP Engine’s Atlas JS framework (which credit to WP Engine is open source).

Developer preference for “modern tools” was mentioned a lot as a big benefit. I’m not so convinced here. Trends in software tools come and go, developers often want to use the next shiny thing, but businesses need to get things done and maintain websites over the long-term. Decisions on your tech stack should not just be down to preferences of your current developers. I’m not a big fan of JavaScript “eating the web” and I think it’s perfectly valid for websites and web apps to be developed in traditional (popular, well tested and maintainable) server-side languages such as PHP, Python or Ruby. If this is you, then WP Engine’s software based heavily on the JS stack is not for you.

Having to force front-end developers to code directly in JS is becoming a big issue in web development (see articles by Chris Coyier, Brad Frost). I think a full JavaScript stack is great if your team can support it. If it can’t then it can be an obstacle to business efficiency and can cause barriers to entry for people coming into the industry.

Front-end development is a highly skilled profession, which includes a wide range of client-side tech and skills such as usability, accessibility, performance, strategic thinking and device compatibility testing. Front-end developers need to know JavaScript, but they don’t necessarily need to be JS programmers.

I find it ironic one benefit of headless is decoupling, yet with the JavaScript/React approach HTML/CSS is tightly coupled with JavaScript. To the point pure HTML/CSS front-end developers can struggle to be able to work with it.

I thought WP Engine’s approach to solving the preview problem in WordPress was interesting. They use OAuth to get users to login to WordPress before they can view preview content, both helping authenticate to the preview site and authenticate the preview API requests to WordPress.

Benefits of exploring headless

Decode hosted a panel discussion on “Navigating the risks to reap the rewards of Headless WordPress” with perspectives from agency 10UP, the creator of WPGraphQL and WP Engine.

Phil Crumm from 10UP noted clients who are most interested in headless are those with a clear functional use case, those interested in security, or those who want more flexibility for the future. Content portability and flexibility is important. “While we don’t know what the future will look like, the decoupled architecture is probably going to be the best way to get there safely.”

Matt Landers from WP Engine talked about when is headless the right approach. His answer was: scalabilty, developer resources, security and data integrations.

He said “do you have the developers resources to do this? It’s a lot more developer heavy. You can’t rely on community built themes and plugins like you can in a traditional WordPress world. You’re going to have to depend on a development team to build that our for you. [maintenence costs] are coming down as we figure out how to integrate the front-end and back-end more seamlessly.”

Jason Bahl from WPGraphQL talked about the benefits of componentisation which could reduce production and maintenance costs in the future. Creating re-usable components that use GraphQL data fragments to output content that can be re-used on projects. I admit I’m not that familiar with how this works in React. We use Timber a lot at Studio 24, an excellent Twig templating system for WordPress from Upstatement. This is another way to support componentisation in WordPress today, though in practise it can he hard to actually do this across different client projects.

On the future of headless in WordPress Jason noted we’re “super early in this game. The more tooling we see built, the more adoption we’ll see.” There seems to be lots happening on this front across lots of different technologies, which is always great to see.

Gutenberg and headless

The final Decode talk I watched was “The Fast Track to Mastering Modern WordPress” with Rob Stinson of WP Engine and Carrie Dils. The talk focussed on Gutenberg and the future of full site editing that is coming to WordPress in 2021.

Full site editing allows users to edit and layout an entire site via Gutenberg, not just the content area but also headers, footers and the sidebar. Carrie went on to explain how full site editing themes works, moving from template files and template parts to “block template parts” which interestingly seem to be straightforward HTML files (which sounds good). It’s not clear to me if all the dynamic content is then in Gutenberg, which splits up all HTML code into React files.

More details on how this works is available at fullsiteediting.com, which notes HTML files may actually be saved as PHP files to support text translation and dynamic URLs. This feels like a pretty major architectural decision. I’d love to see a proper templating engine like Twig or Handlebars in WordPress! It would certainly help developer efficiency and a large breaking change like Gutenberg is the time to do it (if ever).

Carrie went on to show how the editing experience works in WordPress. It’s very similar to editing content in Gutenberg, allowing users to edit areas of the site such as the footer.

As for the timeline of all this, full site editing beta is available in WordPress 5.7 in March 2021. This is expected to be in WordPress core by 5.8 in June 2021, pretty soon! Given the disruption caused by the original Gutenberg release I hope this is a smooth and opt-in transition.

So does the Gutenberg layout need to look the same as the frontend? It’s not clear to me what the WordPress recommendation is on this – but it certainly feels like that’s what Gutenberg wants users to experience.

The obvious issue for headless is it’s decoupled, so laying out a site in Gutenberg cannot easily look the same as the frontend, if you’re using WordPress as a headless CMS. Which, to be honest, I think is good.

A CMS should not have to show content exactly as it will display on the frontend, but should give a good enough representation to the user to allow them to manage content effectively.

Rob Stinson then talked about making custom blocks, an area of great interest to me since I see this as essential for digital agencies taking up Gutenberg. Rob talked about a plugin called Genesis Custom Blocks.

The UI for adding a new custom block appeared to use Gutenberg too. This allows you to setup editor fields (that appear in Gutenberg) and inspector fields (that appear in the inspector sidebar).

I wonder how the custom blocks definition is stored, in the database or filesystem? All projects we use need to be stored in version control so we can deploy them across environments effectively.

When adding a custom block a PHP and CSS file is created which controls how the block appears in Gutenberg (when previewing rather than editing). This certainly makes it easier than embedding HTML in React. This is the same HTML and CSS as used on the frontend.

In the demo Rob noted the “styles on the frontend don’t perfectly match up what’s in the editor, but this can be fixed up.” In our previous experience we found it’s very time consuming, and thus not commercially feasible, to try to match CSS on the CMS backend with the frontend website. This is true for custom CSS themes which are the bread and butter of many digital agencies work. And I would have thought even more true for headless CMS setups where the CSS isn’t even shared.

One great point of the Genesis Custom Block plugin is you do not need to use the Genesis framework to use it. You’re free to use your own HTML and CSS.

We are still concerned about accessibility of the Gutenberg editing interface at Studio 24 and it contributed to our decision not to use WordPress for the new W3C website. I hope the Genesis Custom Block developers have considered this.

The aim of the Gutenberg project is to “make writing rich posts effortless.” That’s a mighty fine aim, but I hope WordPress can achieve this without breaking the web for others (be they accessibility users or front-end devs less comfortable writing HTML templates in JavaScript).

Final notes

Atlas sounds like a really interesting develoment from WP Engine and I welcome more tools in the area of Headless CMS. It’s sure to help more people get involved in this exiciting area of digital technology.

It’s a shame the name Atlas is so close to HumanMade’s Altis, WP Engine certainly would have been aware of this. However, Altis is software that extends WordPress more dramatically with DXP/enterprise features. Atlas, for now, is primarily a solution to help make better Headless CMS sites with WordPress.

Weeknotes: 31 Jan

I’m going to try to start keeping weeknotes. It’s a good way to reflect on the past week, get into a regular habit of writing, and is a nice way of working in the open.

My past week has been pretty busy, in general January has been a pretty crazy start to the new decade with lots going on! Monday and Thursday in London for various client meetings in and around Westminster. We’re currently working with the University of Cambridge on a site for their Alumni magazine and a range of WordPress sites for Parliament.

Had an exciting potential new client call on Tuesday, more on that later!

Had a meeting for the Cambridge Film Trust. We’ve going through a lot of planning for the 2020 festival at present.

I took a look at how to multilingual properly in WordPress using a multisite approach. So far initial work is showing this is a far more sane way of doing it in WordPress. Our previous efforts had included tools such as WPML which we found can get into real difficulties once you start having more complex content.

We’re currently recruiting for a new web developer, so spent some time going through CVs and emails from candidates. It’s a pretty time consuming process, but essential to do properly to get the right people.

Ended the week with a board game night at work, where we played Dixit, a fantastic card game I’d not played before. One person plays the storyteller, chooses one of their cards (with beautiful imagery), picks a phrase, and everyone else has to choose which of their cards best fits. The idea is to guess which card is the storytellers. Will have to buy it for the family to play!

Local development with Valet

I’ve used MAMP Pro with my team for local development for many years. It’s a convenient tool, but it’s often slow (especially the CLI) and often hangs or crashes. I’ve been looking around for alternatives for quite some time, so after it was recommended by Zuzana thought I’d take a look at Laravel’s local dev environment Valet.

Installing Valet

Valet installs a lightweight set of tools to run websites on your Mac via Homebrew. This feels like a good approach to me. I’m not sure we really need the complexity of virtual machines, the code we write tends to work fine on a Mac. Having tools locally installed via Homebrew is fast and convenient.

I started by updating Homebrew to make sure my local packages are up to date:

brew update

Then installed Valet globally via Composer, and installed it:

composer global require laravel/valet
valet install


Next I need to install MySQL since this will no longer be available via MAMP:

brew install mysql@5.7
brew services start mysql@5.7

This has installed MySQL, I now want to secure the local root password (by default it is empty):


(please note the path to mysql_secure_installation may change depending on your version of MySQL).

Next, I want to connect to the database to install a copy of my own blog WordPress database for local testing. We use the excellent SequelPro for managing MySQL.

It’s easy enough to connect to localhost using the details:

  • MySQL Host: 127.0.01
  • Username: root
  • Password: (the secure password I set above)

This worked fine, I created a new database for my local test site and imported a recent SQL backup.

Serving a site

The default method to serve sites in Valet is to use valet park to serve all sub-folders in ~/Sites as websites via URLs in the format folder-name.test

For example, http://wordpress.test/ would serve a website with the document root of ~/Sites/wordpress

This isn’t really how we work, most projects have a “web” sub-folder inside a project to allow for files outside of the document root. So for setting up local sites I’ll need to use the valet link command to set each site up manually.

To create a link from ~/Sites/simonrjones.net/web for the host simonrjones.test it’s simple enough:

cd ~/Sites/simonrjones.net/web
valet link simonrjones

I can verify the sites setup in Valet via:

valet links

The final step is to ensure WordPress knows about this new local test URL, otherwise WordPress has a habit of redirecting to what it thinks is the correct blog URL.

I use the multi-environment config on my personal blog so this is easily achieved by changing the ‘domain’ value for the ‘development’ environment in the file wp-config.env.php

 'development' => [
        'domain' => 'simonrjones.test',

I can test this now via the URL http://simonrjones.test/ – which worked first time!

Testing with a more complex setup

Next, I tried this with one of our client’s WordPress multi-site installs, which is a little more complex. The site is hosted at WP Engine so it uses the standard wp-config.php setup (not multi-environment config).

CD’ing into the project folder and running valet link is enough to make the site run from a local *.test URL. However, trying to access the site doesn’t work and WordPress multi-site redirects the request to what it thinks is the correct URL.

I have WP CLI installed, so I tried to use this to update old site URLs to the new *.test ones, but I couldn’t see a way to reliably do this for multi-sites. So I went back to straightforward SQL!

UPDATE wp_options SET option_value='http://clientdomain.test' WHERE option_name='siteurl' OR option_name='home';

UPDATE wp_site SET domain='clientdomain.test' WHERE id=1;

UPDATE wp_sitemeta SET meta_value='http://clientdomain.test/' WHERE meta_key='siteurl';

UPDATE wp_blogs SET domain='clientdomain.test' WHERE blog_id=1;
UPDATE wp_blogs SET domain='site1.clientdomain.test' WHERE blog_id=2;
UPDATE wp_blogs SET domain='site2.clientdomain.test' WHERE blog_id=3;
UPDATE wp_blogs SET domain='site3.clientdomain.test' WHERE blog_id=4;
UPDATE wp_blogs SET domain='site4.clientdomain.test' WHERE blog_id=5;

UPDATE wp_2_options SET option_value='http://site1.clientdomain.test' WHERE option_name='siteurl' OR option_name='home';
UPDATE wp_3_options SET option_value='http://site2.clientdomain.test' WHERE option_name='siteurl' OR option_name='home';
UPDATE wp_4_options SET option_value='http://site3.clientdomain.test' WHERE option_name='siteurl' OR option_name='home';
UPDATE wp_5_options SET option_value='http://site4.clientdomain.test' WHERE option_name='siteurl' OR option_name='home';

In addition I had to set the wp-config.php setting to the default site URL

define('DOMAIN_CURRENT_SITE', 'clientdomain.test');

Finally, I had to setup additional URLs to serve the multi-site URLs in Valet:

valet link site1.clientdomain.test
valet link site2.clientdomain.test
valet link site3.clientdomain.test
valet link site4.clientdomain.test

The above appeared to serve the multi-site install from the different URLs. However, when I attempted to login to WordPress I was presented with a white screen of death.

I tried valet log and viewed the Nginx error log. Nothing. I refreshed the page a few times and WordPress came up. Navigating around most WordPress admin pages seemed to work, but I got the occassional slow page load or white page. On the front-end pages in sub-folders often seemed to not work on the first request. This is a little disoncerting. This may be caused by the complex WordPress setup, one of the installed plugins, or simply the fact it’s using Nginx and we’ve built these sites to work on Apache. It’s something I’ll have to look into.

Next steps

Valet certainly seems to be a useful tool and one that is very quick to setup. Things I’d like to look at next..

  • Review the secure HTTPS option in Valet
  • Take a look at Valet+, though it looks a little out of date compared with Valet.
  • Can we customise the webserver used so it more closely matches our production environment (we use Apache instead of Nginx FastCGI)?
  • Is there a WP CLI plugin to help change multi-site URLs? If not, we should write one!

If we can use a lightweight dev setup such as Valet it would be nice. Though as with everything in tech, it’s clearly not completely plain sailing! MAMP Pro is certainly easier to use and setup, though it’s the speed (web requests and CLI) which I’m starting to want improved more than MAMP appears able to offer.

Creativity, Playdate and making things

It was with some delight I opened this month’s Edge magazine which I found on my table this evening. This month they have an exclusive on a nifty new handheld console built by software developers Panic called Playdate.

From the article, Playdate looks awesome. It’s a a cute yellow handheld with a high-quality LCD screen, simple controls and an intriguing crank, designed for fun gaming experiences written by indie developers. It looks like nothing I’ve played before. The concept seems pretty crazy, but that seems to be the point.

Edge #333 - Playdate handheld console

With a series of fun, creative, offbeat games released every Monday over wifi once you boot the Playdate up, the concept seems to be genuinely original and born out of a desire to just have fun and make stuff. I can’t wait to get my hands on one!

If you’re a fan of gaming the Edge article is well worth the price of the magazine. Reading through a few things stuck out for me.

I’ve been aware of Panic for many years. We use their software at Studio 24 (their excellent file transfer tool) and I’ve always been struck by their attention to detail and quality of design. I eagerly played through Firewatch when it was released on the Playstation 4. A fantastic game beautifully designed, full of atmosphere and good storytelling. The sort of game I really enjoy.

From the Edge article, Cabel Sasser (co-founder) explains what he believes was the origin moment for the project. He talks about Panic being a 20-person company with revenue around the two million dollar mark. He woke up one morning with “a bit of an existential crisis”. He had a profitable, independent company without external investors, with a team that could put their hand to a range of things – not just the same sort of work they’ve been focussed on for so long.

I realised we don’t have to keep doing the exact same thing that we’re always doing – this ceaseless develoment-and-support cycle. We can do some weird things too, as long as we’re not betting the farm. If we have this chance, we should probably start doing some things that take us to new places. Maybe they’ll work, maybe they won’t. But if we’re not doing that, we’re just wasting out lives.

Cabel Sasser, Co-founder, Panic

This kinda resonates with me. I run a digital agency, not far off the same size and revenue as Panic. I started as a creator too, hacking together web pages and making software. Making things has always been part of my make-up. But with running an agency that often takes a back seat, and spare time pretty much disappears if you’re not careful. It’s fantastic to see other companies of a similar size spin out internal projects into something so impressive.

Another lovely quote is:

Running this company now, I feel almost like a different person. I feel like a huge part of that is finding again how important it is to me and for everyone here to just make things, and be proud and excited about it.

Cabel Sasser, Co-founder, Panic

It’s a bold and exciting move for Panic, but one that I’m sure will give them new opportunities and rewards. There seems to be a real movement for more interesting, playful gaming at present. I hope Playdate does really well.

I’ve seen this in other companies recently too. Only this week I read that the excellent WordPress agency Human Made released Altis, their own “next generation” CMS platform built on top of WordPress. From my brief review it looks like a really interesting set of content and development tools to make creating engaging websites way easier. It sounds like an interesting venture for Human Made.

It also reminds me of Brendan Dawes’s talk at New Adventures about creating things, hacking together technology and making new things. (if you’re not aware New Adventures is a superb conference on digital creativity, ethics, inclusion and other essential topics)

Digital is so powerful and teams that work in this industry end up with such a variety of skills that can be put to great use. Studio 24 is twenty years old this year and I hope to be able to make more time for creating things with my fantastic team. I’ve started already with a foray into building sites with Headless CMSs and some tools I hope we can spin out into a viable open source project in the near future (which I also hope will spark off some interesting talks I can do at user meetups & conferences).

I’m also trying to blog more these days. Blogging on your own site seems to be coming back into vogue, I think it’s simply a nice way to note down your thoughts to help inspire and motivate. As Field Notes neatly puts it: “I’m not writing it down to remember it later, I’m writing it down to remember it now.”

Excellence in entrepreneurship: an audience with Google

I’ve just returned from an excellent talk organised by Cambridge Judge Business School’s Entrepreneurship Centre. Simon Hall hosted a fascinating discussion with Google leaders Jonathan Rosenberg, Senior VP of Alphabet and former Senior VP of Google Products, and Alan Eagle, Director of Executive Communications at Google.

The talk was primarily about the value of coaching, and was celebrating the life of Silicon Valley luminary Bill Campbell (who I admit I did not know), business coach to Steve Jobs, Eric Schmidt and others.

The event was primarily targeted at CJBS students though there was a fair crowd of entrepreneurs and local business types, I chatted to a few after the talk.

Two things happened in the talk I wasn’t expecting. First, quite how entertaining Jonathan and Alan were together. They were clearly good friends and had a hilarious double act going on between them. Alan’s description of how Bill Campbell grilled him in a job interview was brilliant as he made Simon stand in for himself while he channeled an intimidating Bill (there’s a good photo of this moment here).

Secondly, everything Alan and Jonathan said was focussed around human values. How important empathy and developing your team is to business success and how coaching is a great way to achieve that. I wasn’t expecting the talk to be focussed so much on how important human-centered values are to business, but I was very glad to hear so much from the two Silicon Valley business leaders that I found myself agreeing with.  

Alan retelling the story of how he was interviewed by Bill Campbell

A few notes from the talk…  

Interview as much as you can, good to learn how to tell good people.
A good question to ask about experience: What did you learn from this?  

If you have to let someone go do it with dignity. Why? Because it’s the right thing to do. Because it will affect the rest of your team if you don’t. Because it’s a small world, people who you let go may be future opportunities / they may talk about your business.

5 elements of a successful team:

  • Safety
  • Clarity of goals
  • Respect
  • Big mission that matters
  • A meaningful role

Find out more at Project Oxygen  

An effective manager needs to marry the principles of coaching with management.  

Important future tech skills (Alan repeated this a lot):

  • Computer science
  • Machine learning
  • Data

Soft skills:

  • Passion, interest in learning
  • Smart creative dedicated to learning
  • Good communicator, concisely make your point, speak with passion

Important for Google’s success: Speed & simplicity  
All sorts of latency exist so speed really matters. Fast results makes people come back.

The concept of no managers didn’t really work for Google. We had this for 18 months or so. Asked team and they wanted someone to mentor them & take decisions.  

Engineers need a career ladder to rise to the highest level in a company without having to be managers. If you’re the most senior engineer the impact you can have on a tech company is profound & you should be paid the same / more than managers. Not many companies do this.  

Guide & lead, give people freedom. Don’t micromanage.  

There was a question about whether too much growth is bad. Seems not. You can hire brilliant people off the back of fast growth. With an internet business you can often support fast growth. Don’t worry about getting everything right / perfect.
Great quote: “If everything’s going right you’re not going fast enough”    

You can read more in their latest book, Trillion Dollar Coach, available from all good bookstores. I’ll be enjoying the copy I bought this evening!

Adding a Staging environment to Symfony 4

Environments in Symfony

We use Symfony a lot at Studio 24 for building modern web applications. Our normal setup is to have a local development environment, a staging environment for clients to test code, and a production live site.

By default Symfony supports the following environments:

  • dev – (development) intended for development purposes, can be used locally or on a hosting environment to test your application
  • test – (automated tests) intended for use when running automated tests (e.g. phpunit)
  • prod – (production) intended for the live web application

Ideally we would have a third web environment to represent staging, which is what we use to preview functionality before go-live. So that’s what I set out to do.

Adding a custom environment

I want to call my new environment stage to represent staging, since Symfony already uses shortened versions for other environments.

It turns out you can just add any old environment name and Symfony recognises this. So setting the new environment locally is really only a matter of updating your local environment settings file .env.local (you can also set this via actual server environment variables).

# Website environment

Environment configuration

Symfony loads environment variables from .env files. It uses the .env file for default values, then loads the .env.{environment} file for environment-specific settings. Finally, it loads the .env.local file for sensitive variables (e.g. API keys or database credentials – this file should not be committed to version control).

To help keep track of my staging environment variables I created a file at .env.stage to store these.

Package configuration

Different packages use YAML config files in the config/packages folder. I created the folder config/packages/stage/ which is used to store package configuration for the stage environment. It’s possible to inherit values from another environment via the imports key, which is really handy. Here I’m importing the prod settings for the stage environment.

# config/packages/stage/monolog.yaml
- { resource: '../prod/' }

Composer packages

One gotcha is your PHP code may depend on a library that is loaded by Composer locally in your dev environment (via require-dev), but does not load on stage or prod.

When I first tested the above code, it crashed since Monolog was not found (this is used for logging). It turns out Monolog was loaded in my local dev environment via symfony/debug-pack which is setup to only install on require-dev in my composer file.

This simple composer require command quickly fixed it!

composer require symfony/monolog-bundle

Debug mode

Debug is enabled by default for dev and test environments, and disabled for prod and any new environments.

Normally I’d recommend not displaying debug mode for a staging site, since it is supposed to be an environment to preview the site and should work in the same was as production.

However, you can enable debug and the Symfony debug bar for your new stage environment. First set APP_DEBUG in your .env.local file:


Next, ensure the debug bundles are enabled. Edit config/bundles.php and ensure the WebProfilerBundle and DebugBundle are both enabled for the new stage environment.

    Symfony\Bundle\WebProfilerBundle\WebProfilerBundle::class => ['dev' => true, 'test' => true, 'stage' => true],

Symfony\Bundle\DebugBundle\DebugBundle::class => ['dev' => true, 'test' => true, 'stage' => true]

Finally, create the following config files:

# config/packages/stage/debug.yaml
- { resource: '../dev/' }
# config/packages/stage/web_profiler.yaml
- { resource: '../dev/' }
# config/routes/stage/web_profiler.yaml
resource: '@WebProfilerBundle/Resources/config/routing/wdt.xml'
prefix: /_wdt

resource: '@WebProfilerBundle/Resources/config/routing/profiler.xml'
prefix: /_profiler


That’s it! It turns out it’s very easy to setup new environments for Symfony 4, most of the work is in enabling any bundles you require and ensuring the right config files are setup for your new environment.