The Mindset Behind Declarative Programming

Follow the Shit Left motion

Picture from Eric Tabuchi

I felt like a ghost.
Back then I supported the company’s data initiative, partnering with a data platform team.

We weren’t the stars of the company. That’s OK.
Still, I couldn’t explain to myself why nobody seemed to care.

To me, it was where data engineering work shines. Where it scales the most.

Still, I couldn’t put words on it. I couldn’t advocate for the great mindset we shared altogether.

Through the following lines, I finally uncover the mindset behind platform engineering, and how shifting operations to the left makes our work clutterless.

The Shit Left Motion

Shifting operations to the left is about automating every process that sounds like noise. You can find such processes everywhere.

It’s the Tableau to CSV export to Excel import. It’s the manual VM provisioning with the clunked AWS console. It’s the time you copy-paste the hundred times that Google Slide to a new deck because the design team never enforced templates.

These actions should be automated, tested, and versioned to make them explicit. Ultimately, transparent.

Shifting left has some natural challenges: you have to anticipate more. You have to understand end business users. You have to make their work easy.

It’s possible to solve these disturbances. It actually seems natural for lazy engineers.

With this mindset, we relentlessly want all people to shift left. Not for the fancyness. But for the excellence it provides.

In reality, it’s not a place so much as it’s a horizon. There will always be people who are hard to convince, with other priorities, or more important subjects to cover.

We usually fail here by focusing too much on technical aspects. Instead, we should show the value.

Shifting operations to the left limits shadow data operations. It gives better control over users and resources. It makes everyone calm and with a better view of what’s going on.

It takes everyone to the same referential.

Beware, it’s not about scaling everything. It’s about making everything explicit.
Removing any shadow, any corner.
Raising the signal above the noise.
Making things easy so everyone understands the context of operations.

How to make things explicit?

The CEO cares about the inputs and the outputs. He doesn’t know the details of internal processes.
He simply declares what he wants.
Great CEOs have the semantic to give instructions people understand.

We should be like the CEO.
We must have a declarative semantic.

Beware, using a declarative semantic is explicit about the result but not about the process.
Technologies using a declarative interface focus on showing the results. End users declare what results they want.
It’s not all magic.