Saturday, October 04, 2014

The Uselessness of HTML Imports

It is the opinion of this developer that HTML Imports, championed by the Polymer Project and already implemented in the Blink fork of WebKit, are utterly and completely useless.
Beyond this base uselessness, I believe HTML Imports are actively and wildly harmful. I believe the more they are used, the more problems they’re going to cause.

The Case For HTML Imports

Giving the new feature their fair shot, I’ll lay out the case in favor of HTML Imports. I was actually pretty excited about them when I first read the details, and even when I first started using them.
Unique loading of resources without double loading, such as when two imports each themselves import a shared dependency. is really cool! We’re building pretty complicated JS loading systems to solve what, here, is a built-in feature.
The collection of scripts, markup, and styles associated with single units of behavior is really great as a way to conceptualize of how you’re pulling in these web components. It just makes sense to pull in the idea of “Tabbed Pane” and get with it the mark up, behaviors, and default styling for panes of content with tabs to switch between them. No more making sure you have the right scrpt and link tags to match on top of making sure all the places you use the related mark up are correct.

The Case For Not Caring About HTML Imports

You’re already loading Javascript and CSS just fine. Honestly, is it that difficult to include a script and link tag pair? Sure I complained about it above, but let’s be honest, these are small things to complain about.
You already have template loaders if you’re doing any kind of client-side rendering, so HTML imports seem to solve a problem you already have solved. The catch being, they don’t actually solve them because HTML Imports load the documents without actually doing anything with them.

The Case Actively Against HTML Imports

Combining HTML, CSS, and JS into single files isn’t fun to edit or syntax highlight. If you don’t want to hate your editor, your HTML Imports are going to end up being only HTML and they’ll still just use script and link tags to pull in the related JS and CSS.
By introducing these assets externally to keep your sanity, the already asynchronous HTML imports are going to also be waiting on these other assets, and tracking the readiness of these imports gets even more difficult than asset readiness for HTML Imports on their own or CSS/JS loading on their own.
Breaks the value gained from CSS/JS minifiers and preprocessors, because if you’re housing all of these in HTML Imports you aren’t combining them. I’ve seen some proposals that our pipelines learn to extract and combine the JS and CSS from our dependency trees of imports. At that point, what’s the point of HTML imports?
And of course, this increases the number of HTTP requests without clear ways to combine as we would CSS and JS files.

The Case Needed For HTML Imports

Frankly, at this point, I wonder why HTML Imports were introduced at all. What is this really gaining us? What problem did the HTML Import solve? It is really hard to gauge a solution when the problem it aims to help with isn’t even clear in the first place.
If you have an understanding of them that clarifies or counters any of this, please let me know.

No comments:

I write here about programming, how to program better, things I think are neat and are related to programming. I might write other things at my personal website.

I am happily employed by the excellent Caktus Group, located in beautiful and friendly Carrboro, NC, where I work with Python, Django, and Javascript.

Blog Archive