Gary: Iterating on Merge Proposals
Developing new WordPress features as plugins has been a wonderfully valuable process for all sorts of features to come into being, from the MP6 Dashboard Redesign, to oEmbed endpoints, and including multiple Customiser enhancements over the years. Thanks to the flexibility that this model offers, folks have been able to iterate rapidly on a wide range of features, touching just about every part of WordPress.
The “Features as Plugins” idea was first introduced during the WordPress 3.7 development cycle, during which the features were merged after a short discussion during a core chat: it was only in the WordPress 3.8 cycle that the idea of a merge proposal post (called “Present Your Feature” back then) came into being. It was envisioned as a way to consult with WordPress leaders, key contributors, and the wider WordPress community on the readiness of this feature to be released. Ultimately, WordPress leaders would make a decision on whether the feature was right for WordPress, and the release lead would decide if it was ready for that release.
Since then, most feature plugins have published some form of merge proposal post before they were ultimately merged into WordPress, and they’ve nearly all benefited to some degree from this process.
The merge proposal process has worked well for smaller features, but it struggled with larger changes.
The REST API is a great example of where the merge proposal process didn’t work. The REST API was a significant change, and trying to communicate the scope of that change within the bounds of a single merge proposal post didn’t really do it justice. It was impossible to convey everything that was changing, how it all worked together, and what it meant for WordPress.
I’d go so far as to say that the shortcomings of the merge proposal process are at least partially responsible for why the REST API hasn’t seen the level of adoption we’d hoped for. It’s managed to gain a moderate amount of popularity with WordPress development agencies, and a handful of plugins use it in some ways, but it never really entered into mainstream usage in the ways it could’ve.
In a project that prides itself on being willing to try new ideas, the merge proposal process has remained largely static for many years.
Gutenberg is the first opportunity since the REST API was merged where we can examine the shortcomings of the merge proposal process, and see how we can apply the original intent of it to the Gutenberg project’s scope and long term vision.
Over the last six months, Gutenberg project leads have been consulting with teams across the WordPress project. Helping them get involved when they didn’t have any Gutenberg experience, explaining how their focus fit into the vision for Gutenberg, and listening to feedback on where things needed to be improved. In many circumstances, this consultation process has been quite successful: the WordPress Media and REST API teams are great examples of that. Both teams have got up to speed on the Gutenberg project, and have provided their valuable experience to make it even better.
That’s not to say it’s been entirely successful. There’s been a lot of discussion about Gutenberg and Accessibility recently, much of it boils down to what Joe Dolson summarised as being “too little, too late”. He’s correct, the Accessibility team should’ve been consulted more closely, much earlier in the process, and that’s a mistake I expect to see rectified as the Gutenberg project moves into its next phase after WordPress 5.0. While Gutenberg has always aimed to prioritise accessibility, both providing tools to make the block editor more accessible, as well as encouraging authors to publish accessible content, there are still areas where we can improve.
While there’s much to be discussed following WordPress 5.0, we can already see now that different teams needed to be consulted at different points during the project. Where Gutenberg has aimed to consult with teams earlier than a previous feature plugin would’ve, we need to push that further, ensuring that teams are empowered to get involved earlier still in the process.
All feature plugins in the future, great and small, will benefit from this iteration.
Creating a framework for more fluid feedback over the entire lifecycle of a feature project is beneficial for everyone. WordPress teams can ensure that their feedback is taken on board at the right time, project leads gain experience across the broad range of teams that work on WordPress, and projects themselves are able to produce a better resulting feature.
They important thing to remember throughout all of this is that everything is an experiment. We can try an approach, discover the weaknesses, and iterate. We’re all only human, we all make mistakes, but every mistake is an opportunity to ensure the same mistake can’t happen again. Sometimes that means changing the software, and sometimes that means changing the processes that help build the software. Either way, we’re always able to iterate further, and make WordPress fun for everyone.
Developing new WordPress features as plugins has been a wonderfully valuable process for all sorts of features to come into being, from the MP6 Dashboard Redesign, to oEmbed endpoints, and including multiple Customiser enhancements over the years. Thanks to the flexibility that this model offers, folks have been able to iterate rapidly on a wide range of features, touching just about every part of WordPress. The “Features as Plugins” idea was first introduced during the WordPress 3.7 development cycle, during which the features were merged after a short discussion during a core chat: it was only in the WordPress 3.8…