Why WordPress Should Not Enable WebP By Default

I strongly suggest that WP not consider implementing ANY proposal from the Performance Team at this time.

There is a serious lack of leadership, oversight, and proper testing from this team and  I believe many of the proposed changes will do more harm than good if implemented without further investigation.

New Plugin and Proposal

The WordPress Performance team released the Performance Lab plugin with its first set of speed improvements.

The plugin includes:

• A check for WebP support
• Generating and serving WebP images for any JPEG uploaded
• Generating a warning in the Site Health Check dashboard module that OpCache is not turned on at the host
• An audit of autoloaded options
• CSS and JS resource check

The proposal includes:

• Enabling WebP by default in WordPress 6.0

Issues with the New Image Proposal

Making WebP the default served image for any JPEG that is uploaded will cause more issues than it fixes.

Nor has this change been tested under enough variety of conditions to warrant rolling it into the WP core.

An overview of the issues include:

• WebP is not the fastest, nor best format to consider
• Fallback images are required – doubling the disk space and inodes used
• No agreement has been reached for baseline testing of anything in the Performance group
• Extensive testing with existing image optimization methods already in place has not been performed

Voicing My Concern

I have participated in the new Performance Team since Day 1 and have voiced my concerns about the new image proposal on multiple occasions.

I was not alone in my dissent.

And none of those concerns have been adequately addressed.

So, I did my own tests that others can easily replicate to make my points.

If you disagree, fine.

Show me your data and your testing methods so we can replicate.

Test Data to Backup My Claims

See Performance Lab Plugin Image Test Results for the overview of my testing methods and results. 

The full case study documentation is available to my Webmaster Training members and will be made available to the Measurement focus group of the WP Performance Team when that team is reformed and fully functioning.

The Performance Team failed to meet its primary objectives with proposed image changes.

Failed objectives include:

• Thumbnail sizes are not uniformly smaller.
• Retaining fallback images doubles disk space storage.
• Retaining fallback images doubles inodes.
• Only .jpg images are being addressed, not .png.
• Pages do not load faster compared to other image optimization techniques.
• Adequate testing has not been performed to avoid conflicts with other image optimization techniques that may already be in use.

Doubling disk space storage and inodes is a serious problem. 

Hosting resources are limited and site owners have to pay more to increase those limits. Even if hosts gives unlimited disk space, they may impose limits on inodes.

My concern with using WebP images by default is that millions of site owners will only hear that they are getting faster load times and will not be warned about the consequences of suddenly doubling up on limited resources – just like they unknowingly do now with most image optimization plugins.

And, if the Performance Lab plugin is deleted, the extra images are not deleted, nor are those images, nor the originals, sequestered in such a way to make them easy to remove – like what is done by most other image optimization plugins.

WordPress should not dictate what file format is used.

Here’s why:

• Some end users sell their images and don’t care as much about load time as clarity of image.
• New and better image optimization formats are on the horizon.
• Image optimization needs to remain a choice by the site owner to either optimize prior to upload or use a plugin/service to optimize after upload.  

How WordPress Can Improve Image Performance

There are several fundamental areas where WP can make image load speed better, and areas it needs to address with regard to the use of thumbnail images.

• Include established and supported modern file formats all the way through the thumbnails.
• Ensure that non-WP developers are seated at the performance testing table, including hosts, optimization plugin vendors, non-coding site techs who specialize in performance.
• Educate site owners on what happens when:
  • An oversized image is uploaded – it uses a thumbnail, not the original
  • A featured image is auto placed – it uses a thumbnail, not the original
  • An image is picked up for RSS – it uses a thumbnail, not the original
  • An image is included in an email – it uses a thumbnail, not the original
  • What file formats are natively supported all the way through the thumbnail

As you can see, the format of the thumbnail images is radically important. And WebP is not necessarily the best format to use, and is not even supported by all applications that use the thumbnail.

WordPress Needs Real Leadership in the Performance Team

Let’s face it, WordPress is seriously late to the table on dedicating a team to all aspects of site performance.

The key here is to all aspects of performance – both on the code and display sides of a site.

The Performance Team needs a moderator for the live chats. But it also needs two Leads to oversee backend and frontend development and testing so that changes are comprehensive and well vetted.

The Biggest Issue

Developing a measurement standard was 1 of the 21 initial areas of focus proposed when the team formed.

Developing a way to measure performance for both backend and frontend should have been the one and only task for the team.

Without agreed upon standards for setting up a variety of sites under multiple, common conditions, and agreeing upon standards for testing both the code and load speeds, how can any change to WP be deemed a success, much less considered for rolling into core?

We need hard data to back up the claims across a variety of site set ups.

Other Serious Issues

• All site owners who care about speed have already made arrangements to speed up their site load time. 
• A proposal for inclusion in WP core was made prior to even getting all of the bugs out of the plugin, much less enough time to do adequate testing.
• Multiple, strong objections were made to some of the proposed changes by the Performance Team, especially with regard to image optimization. And they were summarily swept under the rug.

What I Haven’t Tested in the Performance Lab Plugin Yet

Other areas of the Performance team with significant workflow include:

• Opcache – host server-side caching
• Notifications in the Site Health check module on the WP Dashboard

I see the backside of hosts every day during client site audits. 

And I can tell you for a fact that way too many shared hosting providers are not setting up their servers for optimal performance.

Many of PHP 7.4 as their default, but Brotli and Opcache are not on.

And the PHP Options are set so high that they are just giving more room for the rats to run on sites that are not properly secured, or for some behemoth security plugins to chew up even more resources trying to combat the currently insane cyber security situation we have with high attacks now – all in an effort to avoid overages limiting the client’s site – yet all of this makes performance for the site, and the server, slower.

Putting this type of change proposed by the Performance Team into the Site Health Check module is going to send non-techie site owners to their host to have things changed that may or may not be in their best interest. 

All that site owner knows is that something is issuing a warning and they want that warning to go away.

I want to test these other options in the Performance Lab plugin on a few of these hosts to ensure it does not create a new, and unnecessary firestorm for both site owners and hosts.

Axe to Grind

Some folks may say that I have a bone to pick with the Performance Team because my suggestions and dissents went unheeded.

Not so.

What I do have is a deep concern for the mess these proposed changes are going to make and having the task of cleaning it up.

My primary income is derived from offering speed and security services to bloggers, along with educating DIY site owners and webmaster designers on both speed and security, as they go hand-in-hand.

I joined the Performance Team to:

• contribute my expertise in these areas
• properly test proposed changes against my client’s perfectly tuned site setups
• get an inside peek at what would be rolling into the WordPress core
• prepare my clients and webmasters for the upcoming changes 

I also was concerned that the team may do things that would bork my client’s sites. And that part is coming true. And THAT alone causes me unbearable stress. 

I also know some folks may ask why I didn’t step up to lead one of the focus groups, primarily the Measurement group.

I’m not qualified to lead a WP team or focus group. Here’s why.

• Even though I regularly read some channels in the Make WordPress Slack group to get a deeper insight into changes, this is my first time to actively participate at this level in a WP Core team. I don’t know how they are run or what the responsibilities of a group PoC (Point of Contact) include. I hoped to learn those things while participating.
• My testing methods are strictly frontend load. Someone with experience in both backend and frontend testing would be more qualified to lead that part of the team. I was hoping to learn about backside testing methods.
• I’m not a developer. I struggled to follow most of what was being discussed in some of the focus groups and I have no way to vet those proposed changes.

I am qualified to be on a performance team.

• I’ve been doing deep case studies for WordPress, themes, and plugins for 15+ years.
• I work with theme and plugin vendors to help improve their wares.
• I teach and work with webmasters and designers who are directly responsible for the speed and security of their client’s sites.
• I help my DIY site owner clients make more money through technical assistance and education and I enjoy long-term relationships with those clients because they see results.

Offers Made and Declined

I did offer to bring my testing methods to the group and offered to help determine and set up various site conditions for testing.

Those offers, along with all others who offered the same, were declined by the Measurement group PoC.

They did make a bit of progress on backend testing setup and something with WebPage Test, which is a frontend tester, but that is about as far as it got. I don’t believe any testing was done, and certainly no agreement was reached on what to test, including the proposal to set up a site with an insanely old page builder and Woocommerce – neither of which represents the majority of site setups.

Get Your Facts First

So, before you spew at me, go read the full 6 month’s worth of the Performance Slack channel meetings for yourself, along with every offshoot group’s Slack channel and see what was offered, how many dissents were expressed, and how little testing was done.

And I do want to make explicitly clear that I was not disrespected in any way by any participant in the Performance Team.
My suggestions and offers were ignored or dismissed – just as those of others who initially joined the team were ignored or dismissed – and most of those folks left.

If nobody’s listening, why continue to speak?

We Need a WordPress Performance Team

Considering how much money a slow site impacts the bottom line of the site owner, we need to make every effort to speed up WordPress.

What we don’t need is a bunch of disjointed projects put on a fast-track to a train wreck due to lack of proper testing and comprehensive oversight.

WordPress powers 455 million websites. And that includes an insanely wide variety of setups of everything from enterprise-level sites on dedicated hosting with server admins who built a custom stack and have custom development on the site, all the way down to newbie bloggers on cheap shared hosting with all manner of plugin and theme combos.

It’s a TALL order to adequately develop WP core to meet all of those needs. It’s an even taller order to test that much variety of setup. But, the effort needs to be made.

I hope to be able to contribute to such a team – that means remaining an unpaid volunteer and giving up billable hours for client work to do so – just like every other team member has to do.

It’s worth it.