Developer Tools Considered Harmful

Back in 1968, Edsger Dijkstra wrote a letter in the journal "Communications of the ACM" [1] in which he argued that “go to” statements were, essentially, an anti-pattern in programming and ought to be avoided:

For a number of years I have been familiar with the observation that the quality of programmers is a decreasing function of the density of go to statements in the programs [sic] they produce. More recently I discovered why the use of the go to statement has such disastrous effects, and I became convinced that the go to statement should be abolished from all “higher level” programming languages […]

I was reminded of this in a recent episode of the Future of Coding podcast [2] in which the hosts discuss this paper and, furthermore, discuss their own (occasionally wacky) “considered harmful” lists. This prompted me to think of my own list of “considered harmful"s, and one which I have encountered the most in recent years: developer tools (henceforth: “devtools”). With some caveats, I have been slowly nurturing the opinion that the whole ecosystem of devtools should be considered harmful.

What is a devtool? For me these are programmes, separate from those directly required to write a programme (like a compiler) whose features are designed to improve “developer experience”. This is a broad category: one might even say that vim, with its text objects and motions, is a “devtool”, and while this is probably true, I’m going to arbitrarily restrict this to programmes which have some knowledge of the specifics of the programming language in which the programmer is working, counting things like LSP clients and servers (and similar functionality within IDEs), code formatters, and so on. There are diminishing returns on more fully defining this category, so that will suffice to illuminate the point I’m trying to make.

I should note that I do use many of these tools: I often use an LSP client in my editor (neovim or helix depending on my mood), and I actually think code formatters are a good way to stop people arguing about formatting (and use one myself: ormolu). Having said that, I really do not think that these tools have any significant impact on my productivity when writing programmes, and this is often the main argument I encounter when programmers either try to persuade me that an IDE really is better, or that the lack of good LSP support rules a toolchain out of contention for a project (and so on). Even (neo)vim, which I love, I could perfectly easily programme without if I had to (I did use VS Code for a few weeks while my laptop was being repaired and it was fine).

In my experience, these devtools, in a large number of cases, have engendered a form of learned helplessness in programmers such that they struggle to conceive of programming without them, and believe that they are necessary for their own productivity. I would go further and posit, in the spirit of Dijkstra, that the ability of programmers to just get on with it is a decreasing function of the amount of time they spend talking about devtools.

A (very) large number of really complex, high-quality programmes were written before these tools existed. Devtools, such as they are, improve the experience but not so much that programming is impossible (or even unpleasant) without them. In that sense, while devtools are helpful and make our lives as a programmers more pleasant, this side-effect induces an over-reliance on them and therefore I would consider them harmful in general.