Anaconda is prohibited at OIST.

Beginning April 1st 2025, installing and using Anaconda is prohibited on OIST systems and on OIST-provided devices.

The reason for this is a license change made by the Anaconda company. Read below for more details, and how you can use a workaround to continue using conda.

What is Anaconda and conda

“conda” is an open source tool for installing (mostly) Python and Python packages.

It was originally made as a way for Windows users to get Python and Python packages. It soon also became popular on MacOS (which does have Python, but usually a very outdated version).

Today it’s also available on Linux and is a popular way to distribute Python programs across all three operating systems. Python has long lacked a standard way to distribute applications, and conda has become one way to fill that need.

Anaconda, the company, publishes the “conda” tool, and manages repositories with open source (and freely distibutable) software; mostly Python but they also provide R, Perl and other languages.

What happened?

Recently, Anaconda have been changing the license that you click to accept when you install anaconda or miniconda.

• The new terms say any organization with 200+ members have to pay a per-user license fee.
• Use in education is free, but research, even at educational institutions, is restricted.

This year they have started suing organizations over non-payment. So far only in the US as far as we have heard.

After backlash, they put out changes to their terms that eased restrictions on academic research. But the restrictions (no sharing, no access for third parties, no embedding) creates a legal minefield.

Our response

We recently (early November 2024) asked OIST legal counsel to review the current terms and the statements and FAQ they publish, to determine the risk for OIST. They sent it on to an external expert. The legal opinion is:

OIST is not legally able to use Anaconda in research.

Now, Anaconda has changed the terms over time. The first moves toward this reportedly happened already four years ago, though we are not able to confirm that. And it’s certainly possible that they will eventually relax the terms enough that it would be possible to use at OIST.

But once a company has started to go down this path, you can’t trust that they will not try it again in a year or two. The only legally safe way to deal with this issue is to not rely on products from the company.

Beginning April 1st 2025 Anaconda is prohibited on OIST devices.

An Anaconda installation is effectively equivalent to having an unlicensed copy of Photoshop on an OIST device, and will be treated the same. Using unlicensed software will be reported, and can be grounds for losing access to OIST computing infrastructure.

common questions

Q: Can’t we pay for an institutional license?

A: No. Apart from the moral and financial questions around paying a gatekeeper for downloading freely available open source code, the actual pricing and terms for organizational users are far beyond what we are able to shoulder.

Q: So can a unit or individual researcher pay for a license?

A: No. As an OIST member, OIST would always be required to get an institutional license.

Q: It is free for education. So it’s OK to use in OIST courses, right?

A: From the license standpoint, yes. But you would be teaching OIST students something that they are forbidden from using in their own research. It’s better to teach them a way that they can use directly in their own work. See below for one suggested solution.

What do you need to do?

In short, first you need to delete any Anaconda or Miniconda installation, and you need to remove any application installed with it. Then, you can reinstall conda through conda-forge and reinstall any applications you need.

What is conda-forge, and how is it different from Anaconda?

“conda”, the application, is open source, as are the packages that you download. The licence terms set by Anaconda only apply to conda when installed from Anaconda; and to using the official Anaconda repository.

Conda-Forge is an independent organization that publishes a lot of conda packages, and also distributes conda (and other tools) completely separately from Anaconda. In general conda-forge packages tend to be more up to date than Anaconda, and works better on some systems.

If you remove Anaconda, then install conda from conda-forge, you are in the clear.

For the most part, anything you can find from Anaconda you can find in the community repos. And third party repositories such as Bioconda are available as channels from conda-forge as well.

Here are the steps you need to take

Remove Anaconda:

1. Get rid of the init scripts in your .bashrc. Run:

   $ conda activate
   $ conda init --reverse --all
   $ conda clean --all

2. Remove Anaconda or miniconda directory with rm -rf :

   # Your installation directory is often your home directory
   $ cd installation_directory
   $ rm -rf anaconda3
   $ rm -rf miniconda

3. delete .conda/ and rename .condarc (so you can later see what settings you had):

   $ rm -rf ~/.conda
   $ mv ~/.condarc ~/condarc.old

4. Delete your installed Anaconda applications. That is usually as easy as deleting the directories where they were installed. If you know you did any changes to them, you might want to move the directories somewhere temporarily so you can refer to them until you have the replacement applications installed.

5. Log out, then in again, just so you don’t have any conda related settings active any longer.

Install conda-forge

1. Go to https://conda-forge.org/ and click “Download intaller”.

2. Pick the installer you need - For Deigo or Saion it’s “Linux X86-64”. Right-click on the link and choose “Copy Link”.

3. Log in to Deigo. Type the following and paste the link to the end:

    $ curl -LO <paste the link here>

This downloads the installer. It is typically called "Miniforge3-Linux-x86-64.sh". This is a shell script (it has an .sh ending). Run the installer with bash:

    $ bash Miniforge3-Linux-x86_64.sh

Follow the installation prompts. The license lists only open source licenses for the code it installs or uses during installation. Accept it (look through the license terms if you’re unsure - it’s not that long), then pick the place you want to install it.

Log out, then in again, and you should have your “conda” command back again as usual.

Other Alternatives

If you don’t want to use conda-forge, you can also use our Python modules. They are not always as up to date as conda-forge (especially on the Saion GPU partition - we’re working on it), but should be well optimized for our systems and come with a fair amount of Python modules already installed. More information here: Installing Python programs

Spack is a package manager for HPC systems and packages a lot of scientific software (not just Python-related ones). We are investigating future use of it for our software installations; meanwhile you can use it yourself to install software you need.