As others have said, support for it is still very limited. But even if there was full support.... would it be smart to use it? How would we do that?
Think about it. A typical JS app written with Node JS modules easily contains dozens, even hundreds of (very small) packages. Do we really want that many requests?
Browserify, Webpack, Rollup etc are so popular because they allow us to bundle many small packages into one fast download. With code splitting we can let the module bundler decide at transpilation time, based on the code our pages are actually using and on some configuration settings, how many bundles to create and what to put in each of them. That way we can write many small packages and serve them as a (couple of) big bundles.
My point is that we should divide our code into packages that work well on a conceptual level, then bundle those packages into bundles that work well on a technical (network) level. If we write our code based on optimum network packet size, we end up sacrificing modularity in the process.
In the meantime, using it will probably only add to the confusion. For example, look at the example on the Edge blog:
import { sum } from './math.js';
Note how they add the extension .js
to the from
string? In Node JS we usually write this as:
import { sum } from './math';
So will the above code also work on Edge? And what about named packages? I fear we will see a lot of incompatibility here before we figure out how to make these paths work across the board.
I would hazard to guess that for most developers, System.import
will remain mostly invisible in the browsers and that only the bundling software itself will start to use it (for efficiency purposes) when it becomes mainstream.