3 ms·
I don't brush aside Node programs just because they're Node. I brush them aside because they usually drag in a few MB of dependencies, and melt my (mid-level) c
by wizzwizz4 6y ago
I don't brush aside Node programs just because they're Node. I brush them aside because they usually drag in a few MB of dependencies, and melt my (mid-level) computer with compilation (often OOM-killing everything else I'm doing on the machine, before dying to the OOM-killer itself) – but for all that, I then need to keep the entire thing on my hard drive because the compilation was mere caching, and hasn't given me an executable; I've still got the runtime overhead of Node, and everything that comes with it.
There are a few Python programs I also brush aside for this reason, though substantially fewer. Virtually every Node project I've seen is a spidery mess of dependencies bringing in dependencies bringing in yet more un-auditable dependencies; the worst Python tends to get is Tensorflow, and it's ready to run immediately (compiling C modules aside – though pip does that at installation time, making that a one-time annoyance for all but obscure C packages).
- efdee 6y agoMelt your computer with compilation? A Node program? More to the point, did you audit Tensorflow? If no, then what's your point to begin with? If yes, what made you conclude that auditing Tensorflow is doable, but usually simple NPM modules are "un-auditable"?
- wizzwizz4 6y agoI didn't audit Tensorflow. But I don't install Tensorflow programs, anyway, because I don't have the resources. The point isn't auditing, though; it's auditability. If it's auditable, then somebody's probably done it – but if it's not, you can't rely on just a spot check of a few dice-picked dependencies.