,
13 tweets,
3 min read
Read on Twitter
There's no such thing as abstraction in web tools and never has been. Every framework claims that it will hide the complexity from you, but it always leaks through. You end up having to understand the framework, its dependencies, and the web primitives they're built on.
Fool me n times (TurboGears, Django, Rails, Rollup, Webpack, etc.), shame on you. Fool me n+1 times, shame on me, I guess.
This doesn't mean that no abstractions work. For most programmers, CPUs, TCP, and many others are leakless. Yes, there are cases where they leak. But those cases are unusual. The web is special: we seem incapable of producing an abstraction that doesn't leak in the normal case.
Agreed: proper abstractions are very rare. You have to think carefully and minimize design gaps. I'm struggling to think of one created in the 1980s or later, but I'm sure there are some. Please don't tell people that your new weekend project is one!
Study the designs of TCP and IP. They're very, very good. From a time before we were distracted by our phones every five minutes.
Everyone thinks that they're the special exception when they're in the throes of repeating the mistakes of history!
Let's make this concrete. An example of what would make a leakless abstraction over the web: no web API errors are possible. All failures occur in the tool's domain, not the underlying web domain. "That's impossible!" No, that's most language implementations before 2010 or so.
Example: you build your new language that compiles to JS. Adopt the following policy: if a user can cause code compiled from your language to emit *any* JS error, that's a bug. It's the same policy that Ruby etc. have: if you can cause a C-level error, it's a bug in Ruby.
If you file an issue where you can show Ruby segfaulting, they will treat it as a bug. There's no discussion about whether it's a bug; it's obviously a bug! But users of languages that compile to JS expect the equivalent of that segfaulting behavior. JS always leaks through.
This is a cute saying, but it says nothing about the boundaries or definition of a "layer". Is the entire web a "layer"? How many pages of spec does that encompass? It's OK if you want to round to the nearest 100,000 for simplicity!
Extremely relevant to this thread:
Don't believe claims that a tool will hide all of the complexity of the web platform from you. Don't feel bad about yourself when it doesn't and you struggle; it's not your fault. Don't tell people that your new tool will save them from the web platform's complexity.
I often talk to people who have been programming the web for ~20 years and have the experience and knowledge to try to fix this, but are so exhausted by the churn and pop culture of "newer=better-than" that the thought of doing it sends them back to bed.
