That document seems to be written by a person who only had a cursory look at conda without realizing that there are options that you could use to avoid these issues.
As Dima said, if they want you to install from source, they should provide a compiler and other tools that you can use to compile sage.
I've added some responses below.
Isuru
> Anaconda very often installs software (compilers, scientific libraries etc.) which already exist on Compute Canada clusters as modules, with a configuration that is not optimal.
Given that they allow singularity, this is moot. binary-pkg, python wheels does the same thing.
> It installs binaries which are not optimized for the processor architecture on our clusters.
Sure. But if you are going to use binary-pkg, you'll get the same issue. Python wheels have the same issue.
> It makes incorrect assumptions about the location of various system libraries.
I don't know what this is referring to.
> Anaconda uses the $HOME directory for its installation, where it writes an enormous number of files.
> A single Anaconda installation can easily absorb almost half of your quota for the number of files in your home directory.
Same issue with singularity and binary-pkg. You can install conda to some other place.
> Anaconda is slower than the installation of packages via Python wheels.
Okay. I don't know why this is a huge problem. Most of your time is spent on calculations anyway. Installation time is a tiny fraction.
> Anaconda modifies the $HOME/.bashrc file, which can easily cause conflicts.
You are asked about this in the command line and can opt out.
> A conda recipe forces the installation of R. This installation does not
perform nearly as well as the version we provide as a module (which uses
Intel MKL). This same R does not work well, and jobs launched with it may die and waste both computing resources as well as your time.
You can use R with MKL from conda-forge. You can even use Sage with MKL on conda-forge. (You can't use MKL with python wheels or binary-pkg)