---
robots: nofollow, noindex
tags: retrospective
---

# Retrospective - May projects

We are going to use the **4Ls** retrospective paradigm. We all need to answer the following questions:

- What did you **like**?
- What was **lacking**?
- What did you **learn**?
- What do you **long** for going forward?

> ### Project Name (Person1, Person2)
> **Like**
>
> **Lacking**
>
> **Learn**
>
> **Long**

### Convergence (Ben, David, Jon, Mak, Xu)

**Like**

* Really love team's willingness to jump in and try new things (how we apply variables, using compose, hooks) and give feedback.
* Fixing the css problem and leveraging the platform with css variables!
* Love unifying the tools between v0 and v7, it's easier to rationalize differences between the 2 when they're on the same tooling and sets us up to be far more interchangeable.
* Love progress towards function components and hooks.
* Regular syncs to help keep on the same page.
* Bounce ideas off each other when we hit blocking points, refine on the spot.

**Lacking**

* Until we PROVE that we're getting better performance and customizability, there's uncertainty. We need to remove the styled props to get faster.
* Need better load balancing across the team to be more productive and to be less blocked by incoming changes.
* Maybe at sync we need to just list out all the random little things we collectively can think of and distribute better
* Need to try the components in real world apps; otherwise we'll miss a ton of things and probably prioritize incorrectly.
* `compose` changes weren't ready and checked in and ended up blocking some of the progress. (Lots of churn)
* Tokens and theming were delayed and we'll need to address in June.
* The sass build pipeline was ready enough to use but still needs updates to be compatible with `useStylesheet`.

**Learn**

* compose unification between v0 and v7 is more than just "use compose" - it also relates to how we handle css, how we process slots, how we mix in unknown props, and how the merged options are processed. We didn't have unification starting this, and much of the time was spent getting this all into a state where we can actually use it.
* IE11 polyfill has blocking issues in its current state, but the issues are being resolved. We will need to fork this and try it in product before asking developers to use it.
* We tried retrofitting `styled` around composed components to preserve some of the backwards compatibility but it ends up making bundle size even worse than before by pulling in all of the tools. We need to just break `styles` props and move support out to a compat import.
* Using react-icons with current code is a pain; we need slots to be able to use them, works great with new code using slots.
* Breaking up larger work items into smaller PRs can help speed up learning and getting feedback.

**Long**

* We need to break the current code in v8 and we need a very solid plan to access the old code after upgrading (through import codemods.) We need this right now with Button before exporting in react-next. Otherwise we risk not bringing perf up or bundle down for default cases.
* Product integration and test app building. We need to use the components we ship. (Let's try to do this in Teams/ODSP)
* Customer documentation for IE11, compose and theming
* Release notes for v8 so customers know what to expect and how to try it out.
* Design docs to be updated, and publicly hosted on the component documentation
* Enough momentum to build new components!

### Shield (Elizabeth, Aneesha)

**Like**
- Having PM help with triage was great! Made shield feel much more manageable.
- [Template responses](/wrwvpTX3RjO0AJJJt-6Y5w) are super helpful (at least as a starting point) since I tend to be super slow with writing

**Lacking**
- Still some lack of clarity about where to go with certain types of issues (design, a11y, components contributed by other teams that we don't know as much about)

**Learn**
- It's good to have dev/PM triage together at the same time because you can bounce thoughts off each other and it helps accuracy

**Long**