date: 2020-08-01 layout: post title: pkg.go.dev is more concerned with Google’s interests than good engineering
tags: [go]
pkg.go.dev sucks. It’s certainly prettier than godoc.org, but under the covers, it’s a failure of engineering characteristic of the Google approach.
Go is a pretty good programming language. I have long held that this is not attributable to Google’s stewardship, but rather to a small number of language designers and a clear line of influences which is drawn entirely from outside of Google — mostly from Bell Labs. pkg.go.dev provides renewed support for my argument: it has all the hallmarks of Google crapware and none of the deliberate, good engineering work that went into Go’s design.
It was apparent from the start that this is what it would be. pkg.go.dev was launched as a closed-source product, justified by pointing out that godoc.org is too complex to run on an intranet, and pkg.go.dev has the same problem. There are many problems to take apart in this explanation: the assumption that the only reason an open source platform is desirable is for running it on your intranet; the unstated assumption that such complexity is necessary or agreeable in the first place; and the systemic erosion of the existing (and simple!) tools which could have been used for this purpose prior to this change. The attitude towards open source was only changed following pkg.go.dev’s harsh reception by the community.
But this attitude did change, and it is open-source now^1, so let’s give them credit for that. The good intentions are spoilt by the fact that pkg.go.dev fetches the list of modules from proxy.golang.org: a closed-source proxy through which all of your go module fetches are being routed and tracked (oh, you didn’t know? They never told you, after all). Anyway, enough of the gross disregard for the values of open source and user privacy; I do have some technical problems to talk about.
One concern comes from a blatant failure to comprehend the fundamentally decentralized nature of git hosting. Thankfully, git.sr.ht is supported now^4 — but only the git.sr.ht, i.e. the hosted instance, not the software. pkg.go.dev hard-codes a list of centralized git hosting services, and completely disregards the idea of git hosting as software rather than as a platform. Any GitLab instance other than gitlab.com (such as gitlab.freedesktop.org or salsa.debian.org); any Gogs or Gitea like Codeberg; cgit instances like git.kernel.org; none of these are going to work unless every host is added and the list is kept up-to-date manually. Your intranet instance of cgit? Not a chance.
They were also given an opportunity here to fix a long-standing problem with Go package discovery, namely that it requires every downstream git repository host has to (1) provide a web interface and (2) include Go-specific meta tags in the HTML. The hubris to impose your programming language’s needs onto a language-agnostic version control system! I asked: they have no interest in the better-engineered — but more worksome — approach of pursing a language agnostic design.
The worldview of the developers is whack, the new site introduces dozens of regressions, and all it really improves upon is the visual style — which could trivially have been done to godoc.org. The goal is shipping a shiny new product — not engineering a good solution. This is typical of Google’s engineering ethos in general. pkg.go.dev sucks, and is added the large (and growing) body of evidence that Google is bad for Go.