I do not really understand all this emphasis on using juliaup to install Julia. I will give my two cents on this issue, but I should start with a disclaimer: I am a user, not a developer. I have been teaching a course using Julia and Pluto notebooks since early 2021, with more than 400 students each year across two semesters. The students are not from computer science; they are from economics, finance, and management, and their interaction with computation is very limited. Therefore, in this context, the installation procedure is a critical issue, and our experience may serve as a useful example for the Julia community.
In the first 6/7 semesters, students were told to install Julia manually, and we had no problems with those installations. Not a single one! Students just loved it: it takes less than 1 minute, is effortless, and requires no knowledge of computation (apart from the Linux case; in Windows, just click an executable and fill in a square to “Add Julia to PATH”; and in macOS, it is even more straightforward). Most students have Windows (around 75%), macOS represents the remaining majority, while Linux is a residual part of the distribution. Then, in one semester, following the Julia mainstream, we changed our strategy: we advised students that it was up to them to install Julia manually or using juliaup, but the latter was the “officially” recommended one. That proved to be ill-advised because we had some cases where the juliaup installations were broken. Moreover, those problems became unbearable because we could easily fix a manual Julia installation, but we had no clue how to do the same with a juliaup installation.
Understandably, we have since returned to the manual Julia installation policy. We have never had a single problem with this approach to installing Julia on an already large number of students’ computers (around 2500 by now), many of whom know very little about computers and computation. In fact, the only occasion we had problems with the Julia installation was when we recommended juliaup, and students followed our advice.
For some Julia users and developers, juliaup is a wonderful installation tool. But not for all users (most users, I believe), not compulsory alongside specific IDEs, and certainly not an acute problem in the present state of Julia development. Currently, my colleagues and students complain about Julia, but their concern is not the installation process itself, which they love because it is as easy and fast as it gets; it is about how long it takes to load a Pluto notebook compared with similar notebooks in other languages. For example, to load a standard notebook (reading some data files, importing some images, using some markdown code, performing some numerical computations, and producing some plots), it takes around four times longer in Pluto than in a marimo notebook (marimo notebooks are the Python counterparts of our Pluto notebooks). Yes, the exact same content in the two notebooks produces such absurd differences, considering that Julia is advertised as a fast language. And the point is that this situation seems to be getting worse with the new Julia versions, not better.