What's the point of Mise over ASDF or Aqua?

[airtonix]

So far having used asdf for many years my two main gripes are:

• that it doesn't work on windows.
• it doesn't help me protect against supply chain attacks by:
  • hash pinning my asdf plugins per project. (asdf-plugin-manager tries to address this but it's broken)
  • hash pinning the tool versions

So after using Mise for a couple weeks :

I love:

• it's friendly output
• it leans on the asdf plugin ecosystem

One thing that Mise does which I had always made a asdf wrapper for:

• it auto installs asdf plugins

but I now realise there's no real point to it because what we really need is:

• plugin system that works on bash/zsh/fish/ash/dash/powershell/nushell (basically it can't be based on shell scripting)
• SBOM to protect against supply chain attack
  • asdf-plugin-manager tried to do this, but there's a few issues with it.
  • asdf plugins themselves don't universally support gtisha, and lean on git tags which can be moved.

I have been keeping an eye on another Tool Manager called Aqua that solves all these problems.

We don't really need a task runner or a env manager. There are existing tools that solve these problems properly.

[mcallaway]

I'm pretty new to mise, and this question is how I discovered aqua, so I went over there to compare. Before that, however, I just wanted to comment to the original poster that I do want a task runner so that I don't need Makefiles, and I do want environment variable control. Looks like Windows support is coming to mise (or is already here, to some level). So perhaps hash pinning is your only missing feature?

By contrast aqua says one must use asdf for language support and I'm primarily interested in python. But I don't want two systems (aqua and asdf).

So it seems for me, mise is more feature complete. I went to ask what's the point of aqua. Not yet sure what I think overall.

[airtonix]

@mcallaway cant we link to the ticket tracking this statement?

Looks like Windows support is coming to mise

[jdx]

If aqua solves your problems then you should use aqua. I'm not very familiar with the project but my understanding is that it has more of a focus on security and reproducibility where I have more of a focus on ease-of-use and features.

I've moved mise away from being a simple asdf replacement and it now has support for running vfox plugins as well. I have not investigated this but I could see a backend being added to support aqua through mise—perhaps via a mise-aqua plugin. The problems I'm interested in are more to do with the front-end of dev tools over the back-end so leveraging work that others have done in that space may be useful.

[airtonix]

The vfox plugin support is a big step forward. ❤️

If you tackle the SBOM question, then mise becomes a very strong tool. Supply chain attack is a very real problem that too many people ignore (thus becoming a systemic issue).

[borissmidt]

In my opinion it might be a good idea to implement aqua registry support directly into mise and make it the default backend.
It is easier to register a new too into aqua than asdf it is cross platform compatible and its easier to download the tools in parallel.

I had a few 'broken' asdf plugins which required jq so then i had to make sure jq was also in my tool-versions file.

It might not be that much work if this issue gets resolved:
aquaproj/aqua#3074

[boris-smidt-klarrio]

@jdx Cool that you implemented the aqua backend 👍 i will try it out today.

[powerman]

We don't really need a task runner or a env manager. There are existing tools that solve these problems properly.

Can you name these tools? I've checked a lot, and first one which actually solve these problems properly is Mise.

From top of my head:

• Whole dotenv family of tools are broken by design just because there are no common dotenv specification and you can't use only one of them (and thus stick to dotenv variant of that tool): a lot of 3rd-party tools use their own libraries to read dotenv, do that by default, do that using different dotenv syntax/features, and this guaranteed to give you unpredictable results sooner or later.
• Taskfile in about 7 years since releasing 1.0.0 didn't solved an issue with buggy and counter-intuitive precedence for env vars - which to me looks like a critical issue which I've noticed in a hour when I start playing with Task. A couple of months ago they tried to fix it as backward-incompatible experiment. 7 years, Karl! How many similar issues are lurking there I just don't like to know, sorry.
• ASDF - well, everyone here know why Mise is much better.
• Aqua… it's mostly looks good - for what it does. But it does not install as much tools as Mise (I'm using more than hundred actually, and Mise installed all but 3, while Aqua installed about 25 or so), it does not implement env vars and tasks. Also it shims are much slower than Mise (50 ms vs 8 ms on my system). The only thing it does really good is security - I hope Mise eventually will have similar features.

[jdx]

Whole dotenv family of tools are broken by design just because there are dotenv-linter/dotenv-linter#805 and you can't use only one of them (and thus stick to dotenv variant of that tool): a lot of 3rd-party tools use their own libraries to read dotenv, do that by default, do that using different dotenv syntax/features, and this guaranteed to give you unpredictable results sooner or later.

I hadn't considered this but just by putting this in toml and not .env has a lot of benefits by establishing our own rules about the file. xkcd 927 and all that—I don't see that as valid criticism. Put your env vars in toml and it works, that's what matters.

Taskfile in about 7 years since releasing 1.0.0 didn't solved an issue with buggy and counter-intuitive go-task/task#1038

lol that is ridiculous. I can't imagine anyone would want it to behave the way it does! If that was my project I'd just declare it a bug and flip the behavior—breaking change be damned.

I suppose the answer to OP's question though is that with me at the helm you are more likely to experience change. asdf hasn't had a new feature in 2 years. While I try to ensure stability, my top priority is making mise the best tool manager it can be and when those goals are at odds I lean towards making mise better.

[jdx]

I just discovered just has the same exact behavior as Taskfile: casey/just#733

this is crazy to me. I don't know why anyone would think that's how it should behave. And given these are some of the top issues on both projects—and nobody has asked for this behavior in mise to be different AFAIK—I'm inclined to think I made the right decision with mise.

to me env vars are what was there before you ran the command and since you're running the command, things it sets should overwrite what was there before