Polymer 1.0: element scoped CSS via <link> leaks outside the element (not truly scoped)

[viridium]

I put the picture-frame element, from [1], in a Chrome extension; I included the index.html. After moving the script to an external file to bypass CSP issues [2], it worked fine. Next, I opened index.html and wrapped the content in a div. Then, I went back to the picture-frame element and moved the CSS into a separate file, naturally replacing the style element with a link element.

The style rules suddenly started to apply, incorrectly, to the div from index.html. I tried moving the style after the template element, as per a suggestion from Polymer 0.5 [3], but there was no change. The styles should apply only to the element defining them, whether they are delivered via style or link href.

[1] https://www.polymer-project.org/1.0/docs/start/quick-tour.html#compose-with-local-dom
[2] https://code.google.com/p/chromium/issues/detail?can=2&start=0&num=100&q=&colspec=ID%20Pri%20M%20Stars%20ReleaseBlock%20Cr%20Status%20Owner%20Summary%20OS%20Modified&groupby=&sort=&id=393307
[3] #779

[sorvell]

Including a stylesheet for an element definition via <link rel="stylesheet"> is not supported in Polymer 1.0. The platform currently lacks support for doing this.

Polymer does support a couple different wants to use style data separated from an element's dom-module. See https://www.polymer-project.org/1.0/docs/devguide/styling.html#external-stylesheets which will soon be updated with this info: #2229

[viridium]

Hmmm; interesting. Will it be supported at some time? If not, I suppose I could pay a small overhead cost by wrapping my CSS in some boilerplate to keep it separate from my element's markup -- essentially, I'll make a second CSS-only element next to the main one.

[sorvell]

No, there are no plans to directly support <link rel="stylesheet">.

Because of Polymer shims style data, it needs access to style text. Unfortunately, this is not possible using link. In Polymer 0.5, link was supported via xhr'ing the style data; however, this caused other problems by introducing significant lifecycle timing complexity. We've chosen not to do this for 1.0.