Re: [webcomponents] Interaction of shadow DOM, imports, base URIs

This is a really good question. Web developers expressed the need for
properly resolving URLs pretty much as soon as they heard of HTML

What they want is for this to Just Work Like Stylesheets (tm). In
other words, when they define an import "300.html", then in that file,
they want to use relative URLs to refer to resources:

<link rel="import" href="spartans/dilios.html">
<element name="bottomless-pit" extends="textarea">
   <link rel="stylesheet" href="themes/super-dark.css">
   <script src="weapons/dory.js"></script>

Since custom elements are currently completely decoupled from HTML
Templates and Shadow DOM, instantiating a shadow tree from a template
that came from an import is the responsibility of the custom element
author. The code I've seen in the wild is:

<element name="king-leonidas">
<template><img src="this-is-sparta.png"></template>
var template = document.currentScript.querySelector('template');
    readyCallback: function() {

Here, cloneNode will simply copy attributes, and destroy any
information about import's URL. and will definitely not give the
Spartans^H^H^H^H web developers what they are expecting. Unless:

1) we run import scripts in a separate scripting context that defines
"document" to mean something different, or somesuch.

2) we provide some explicit APIs for resolving relative hyperlinks on
elements coming from one document to another

3) we add a processing step of some sort to HTML Imports, where we
resolve hyperlinks at the time of import.

4) <your idea here>


On Fri, May 24, 2013 at 10:15 PM, Boris Zbarsky <> wrote:
> Basic question: what should the base URI be for a node that's part of a
> shadow tree that comes from some URI that's not the document URI?
> Making it be the base URI of the resulting ownerDocument is nice and simple
> and involves no spec changes and means that relative URIs to things like
> stylesheets or images are totally unusable in the shadow DOM if it's meant
> to be included in documents that have different base URIs... and so are
> absolute ones, really.  This is the behavior Mozilla's XBL has right now,
> but it's pretty busted, in my opinion.
> So what do we actually want here?
> -Boris

Received on Tuesday, 28 May 2013 18:28:55 UTC