Skip to content

Latest commit

 

History

History
81 lines (59 loc) · 3.28 KB

the-v3-curse.md

File metadata and controls

81 lines (59 loc) · 3.28 KB
Raw
title description date
The v3 curse
Breaking changes in software – Too much, too late, too soon
2024-09-13

The v3 curse

Too much, too late, too soon

I once discovered that v3 is where projects go to die. Winamp (for me) died in the v2 to v3 transition, so did Python.

I didn't find a 3rd example, but I thought maybe that was a good omen, cautiously eyeing the number 3.

That discovery was both incorrect and correct. Python did not die. It caused pain and suffering to all involved over a totally avoidable breaking change, but it survived. Vue did the jump from v2 to v3 successfully. And I never found a third example.

But it is also correct. It is not number 3 that has any significance. But there is a curse indeed, the curse of too much, too late, too soon.

Like all coins, there are two sides to this one too.

  1. Most mutations kill an organism. Revolutions rarely succeed, and are painful to all involved.

  2. Sometimes the pain is necessary. And evolution doesn't work without death.

So there is no one right or wrong answer. Depending on the circumstance, sometimes breaking changes are necessary, sometimes they are not.

But it is not just the judgement call (or the lack of it) which is important.

Of equal effectiveness in quelling the curse is the pathway from the old town to the new. Is the new town too far away (too much)? Is the old town already too big (too late)? Is the migration gradual or involved (too soon)?

I came across an interesting stat today (that is indeed what precipitated writing this rant). These are the package managers seen by npm, the JavaScript package registry (source):

  • 64% npm (all versions)
  • 25% yarn v1
  • 9% pnpm (all versions)

Yarn 1 (aka yarn classic), accounting for 25% of all traffic, is unmaintained. The team has been working on a newer, backwards incompatible, yarn for years. Go to their website, and you'll have to poke around to even find instructions how to install yarn classic. Yet it still accounts for almost all the traffic that comes via yarn!

Gazing at those numbers, it seems yarn is another victim of the v3 curse.

Someone trying to transition their existing codebase from yarn classic to yarn not-classic is bound to wonder - if I'm spending the effort to migrate, do I even need yarn anymore? In the years that the yarn migration has dragged on, npm has caught up, and there are new players in town.

I like yarn. I remember the breath of fresh air it felt back them, and I still use it. But I've realized that it is infrequent to have projects where the choice of the package manager is a functional issue, and using anything that is not npm comes with a tradeoff of asking each person using the project to install another dependency.

This is going to be even more onerous now given that, ironically, a PR to enable corepack by default is going to lead to the removal of corepack altogether.

I don't have a conclusion here. These are just thoughts that occurred to me seeing the corepack saga unfold. I don't think npm is going to be the end all. There will be new runtimes, new package managers. But I do I wonder if yarn has done the puts-stick-in-bicycle meme to themselves.