I actually hope this doesn't take off (or at least, if it does, that it keeps its scope in check). RVM is invasive and moderately broken because of that.

And besides, in a very short time, Go is going to commit to API stability with Go 1. And given that, and static typing (even at runtime), and Go's beautiful module system, and the fact that static linking is the norm, all means that worrying about which Go version you are running on is a silly concern.

This tool purports to solve dependency hell, but dependency hell, especially in a language with no stable ABI where everything is compiled from source, is entirely about package management. Go's standard library used to be in flux, which made pinning a version an important piece of ensuring source compatibility with packages. That's about to go away, and I hope GVM-like tools will follow.

And yet look at how many people use RVM. Just from user popularity alone it's clear that system package managers, no matter how much they're praised by sysadmin people, are not adequate. There's a clear need for package management that's:

1. cross-platform.

2. allows users access to the latest version of the software immediately.

3. allows installation of multiple versions. API/ABI stability is great but perfect backward compatibility is never possible. E.g. some apps actually rely on bugs in the platform and if you fix them they break; yeah yeah I know the software must be fixed but there may be a ton of reasons why that doesn't happen or at least not happen immediately and there's a clear need among users to be able to install multiple versions of a platform/library.

RPM/DEB theoretically support (2), but in practice distros never do that. Third party packagers that immediately provide the latest versions are most exception to the rule because they need to replicate their effort multiple times over multiple platforms thanks to the lack of support for (1). RPM/DEB don't even try to support (3).

Maven.

* Cross-platform.

* Easy access to remote packages as soon as they're posted.

* Versions must specified, so builds are very repeatable. You bear the responsibility of updating your dependencies, but you don't get surprised if a version decides to break compatibility, either.

Also, very decentralized, and it's easy to create a directory locally to hold the dependencies within a project (specify a file:// repository relative to ${project.dir}), or on a per-system-user level (mvn install), or on a per-company/organizational unit basis (self-hosted repos are a cinch).

It's actually a fantastic package management system tailored to Java that works extremely well. Its main problem is that it's also fantastically verbose, and including more and more build plugins can grow exponentially more arcane the greater the number of plugins involved. However, with Go's standardized directory layout among other things (I admit, I haven't done much research into trying Go out), something akin to Maven may actually be a good fit.

As a user, i've been occasionally fristrated with the decentralized nature of Maven. The isn't any single repository which would contain all packages (especially rc/beta versions). And in my experience repos and which repos are used to distributed a package tend to change fairly often. More often than not, I have to hunt for new repos when building from scratch.

Maven is also very slow when your project has ~10 repos and you need to download multiple dependencies. Having your own repo might be a necessity.

You can set up Artifactory on a nearby server you own and proxy all of your package requests through it. Works quite well.

Maven is adequate as a dependency resolution tool, though I always marvel at the insane number of packages it downloads.

That said it should not ever be used as a build system -- ant exist for exactly that reason or you could use Make if you want your builds to go faster.

They take opposite approaches. Maven gives you scaffolding; it's up to you to customize it. Ant gives you practically nothing, but you can do whatever you want. If you want to code on the shoulders of giants then Maven will help you much more than Ant will, IMO.

Because Go builds statically linked executable files, the GVM tool only applies to development environments.

As far as the development environment is concerned, I find it easy to manage multiple versions of Go by setting $PATH and $GOPATH as needed. I only work with a couple of versions of Go at a time and I don't share packages between versions of Go, so my situation might be simpler than what they have at moovweb.

I expect that I will mostly use a single version of Go after Go 1 is released next week.

I kind of think of RVM as a necessary bug due to the unmanaged span of Ruby versions.

I am not clear why this is really necessary for GO. Why not just use 'go fix'?

For simple projects or toy apps, gofix is just fine. However, some companies are running complex production systems written in GO 0, and the code may have a lot of dependencies and custom build scripts. In this case it's useful to have an easy way to install and pin Go versions.

Once Go 1 launches, subsequent releases should be compatible with the Go 1, which reduces the need for something like GVM.

> Once Go 1 launches, subsequent releases should be compatible with the Go 1

Which does not mean that they will be, despite Sun's effort to never formally deprecate (let alone remove) anything and the glacial pace of the language, there have been backwards compatibility issues in the past: reliance on implementation details or under-specified behavior which got changed, reliance on bona-fide bugs which just happened to do what a developer wanted in that case, etc...

I am getting that the GoFolk are building GOLANG with this problem firmly in mind, and may just be expecting to avoid it.

You can expect the unexpected not to happen all you want, reality won't care a bit.

In fact, this makes me think of the Douglas Adams quote sitting right above my main display:

"The major difference between a thing that might go wrong and a thing that cannot possibly go wrong is that when a thing that cannot possibly go wrong goes wrong it usually turns out to be impossible to get at or repair"

I wouldn't consider it "necessary" however even the casual Go user can benefit from the automatic environment setup GVM provides. Go is quickly becoming my language of choice and GVM has spared me the hassle of GOROOT and GOPATH setup. Plus I keep anxiously checking `gvm listall` for the Go1 tag

Yeah, I think RVM also kind of proved that build tools that inject themselves via .bashrc are more trouble than they're worth. It was somewhat forgivable with Ruby because they had decades of legacy tools to deal with; one would hope Go would be able to find a better solution.

Sounds like`virtualenv` for Go. But shouldn't the need for this go away over time as Go 1.0 use becomes widespread?

This is slick. I will note though, if you're just getting started with Go or if you only have small projects, this is overkill. `go fix` and just starting out with current weeklies (RCs for Go1) is sufficient. For projects of any substantial size though, this looks like a nice handy tool!