[yongjik]

The flip side is that services with an immediate need will get upgraded, and others won't, and six months later you will be saying "Why am I still seeing this bug in production, I already fixed it three times!"

Of course, the problem can be mitigated by a disciplined team that understands the importance of everybody being on the same page on which version of each library one should use. On the other hand, such a team will probably have little problem using monorepo in the first place.

Whether you have a monorepo or multiple repos, a good team will make it work, and a bad team will suck at it. But multiple repos do provide more ropes for inexperienced devs to tie themselves up, in my opinion.

[spion]

I don't think that's quite true. In my experience multi-repos have the edge here.

If you have one key dependency update with a feature you need, but you need substantial code updates and 80 services depend on it, that may be impossible to pull off no matter what. Comparatively, upgrading one by one may not be easy, but at least its possible.

The importance of everyone being on the same page with dependencies might just be a limitation of monorepos rather than a generally good thing. Some services might just not need the upgrade right now. Others may be getting deprecated soon, etc.

[rfoo]

There are languages / runtimes where there could not be two different versions of the same thing in one binary (and they eagerly fail at build time / immediately crash upon run). That is not the case for JavaScript, Rust, etc. But it is the case for C++, Java, Go, Python and more.

Everyone claims different needs if they can. Nothing could be linked together anymore if you just let everyone use whatever they want.

Or maybe people start to try to workaround this by ... reinventing the wheels (and effectively forks and vendoring) to reduce their dependency graph.

There is a genuine need for single instance of every third party dependencies. It is not unique to monorepos. Monorepo (with corresponding batch change tooling) just make this feasible, so you don't hear about this concept for manyrepos, and mentally bind it to monorepo.

[pjmlp]

> But it is the case for C++, Java, Go, Python and more.

It certainly isn't for Java, hence why multiple classloaders exist.

For C and C++ it depends on the OS, on Windows (AIX, and similar OSes) this isn't an issue thanks to how symbol visibility works.

Two different libraries are free to have whatever versions they feel like.

[rfoo]

Thanks. I'm not familiar with Java. I thought multiple classloaders are more like dlmopen (which doesn't help much - symbol visibility is hard) cause I saw people struggling on classpath conflict etc.

[pjmlp]

It is basically how application servers got implemented, every EAP/WAR file gets their own classloader, and there is an hierachy that allows to override search paths.

That is how I managed back in the day to use JSF 2.0 on Websphere 6, which officially did not had support for it out of the box.

[spion]

> There are languages / runtimes where there could not be two different versions of the same thing in one binary

But I'm not talking about one binary here. I'm talking about multiple, separate services.

[rfoo]

How many internal libraries does your "separate services" contain? You service A depends on library alpha@1, your service B depends on library alpha@2. All happy now. Introduce another layer, your service A depends on library alpha@1, beta@1, and alpha@1 depends on gamma@1, beta@1 depends on gamma@2, what to do now? It does not even matter how many services you have now.

With Javascript it does not apply, alpha@1 can have its own gamma@1, beta@1 can have its own gamma@2. But the same does not hold for most languages.

left-pad is both amazing and sad. It's amazing because JS's "bundle entire dependency closure" approach, combined with npm infrastructure, successfully drove the usability of software reuse to the point that people even bother to reuse left-pad. This is beyond what a well-regulated corporate codebases can achieve (no matter strongly encouraged single instance or not, not matter manyrepo or monorepo), and it happens in open. It is sad because without being regulated people tends to do so too aggressively, causing, well, left-pad.

[spion]

> How many internal libraries does your "separate services" contain? You service A depends on library alpha@1, your service B depends on library alpha@2. All happy now. Introduce another layer, your service A depends on library alpha@1, beta@1, and alpha@1 depends on gamma@1, beta@1 depends on gamma@2, what to do now? It does not even matter how many services you have now.

Got several thoughts on this one. First, lets look at how bad the issue really is:

To start using beta@1 you need to upgrade alpha@1 to alpha@2 that depends on gamma@2. What's the problem with that?

The same situation can arise with 3rd party dependencies, except there its much worse: you have zero control over those. Here you do have the control.

Now lets look at what this situation looks like in a monorepo: you can't even introduce gamma@2 and make beta@1 at all without

1. upgrading alpha@1 to alpha@2

2. upgrading all services that depend on alpha@2

3. upgrading all libraries that depend on gamma@2

4. upgrading all services that depend on gamma@2, if any

So you might even estimate that the cost of developing beta@2 is not worth it at all. Instead of quasi-dependency-hell ("quasi" because your company still controlls all those libraries and has power to fix the issue unlike real dependency hell) you have a real stagnation hell due to a thousand papercuts

My second comment is about building deep "layers" of internal dependencies - I would recommend avoiding it for as long as possible. Not just because of versioning, but because that itself causes stagnation. The more things depend on a piece of code, the harder it is to manage it effectively or to make any changes to it. The deeper the dependency tree is, the harder it is to reason about the effect of changes. So you better be very certain about their design / API surface and abstraction before building such dependencies yourself.

Major version bumps of foundational library dependencies is an indication that you originally had the wrong abstraction. No matter how you organize your code in your repos, its going to be a problem. (Incidentally, this is also why despite the flexibility of node_modules, we still have JS fatigue. At least with internal dependencies we can work to avoid such churn.) It should still be easier with separate services, however, as you can do it more gradually.

Last note on left-pad and similar libraries. They are a different beast. They have a clear scope, small size and most importantly, zero probability of needing any interface changes (very low probability of any code changes as well). That makes them a less risky proposition (assuming of course they cannot be deleted)

Those are my (hopefully nuanced) 2 cents.

[rfoo]

> To start using beta@1 you need to upgrade alpha@1 to alpha@2 that depends on gamma@2. What's the problem with that?

The problem is the team maintaining alpha does not want to upgrade to gamma@2 because it's an extra burden for them, and they don't have an immediate need.

The debate is not about teams owning separate services, it's about teams owning libraries.

[spion]

Note: steps 1-4 will likely be done in somewhat reverse order (3,4 first, then 1,2, then building `beta`)