Hmm. Unsure if this should be specified in CSSOM, or somewhere else.

This will have to be part of CSS' fetching model. Which I guess is still not really defined. But e.g. it also affects the definition of <style> in HTML.

Can you elaborate? I'm not sure I really understand what this bug is about.

The idea is that when you fetch <img src=svg>, you get back a resource that you parse into a document without a browsing context or settings object. You still want to apply style sheets from that Document to create some kind of rendering output (SVG hasn't really defined all this yet unfortunately, so I've no idea what terms we want). However, we do not want that document or its style sheets to fetch any external resources, so they need some additional processing flags for that (that they can then pass on to underlying systems such as Fetch, so that blob and data URLs still work).

To be fair, the apply a style sheet (with some parameters) to a document and get some rendering output, is probably a missing concept in CSS, not SVG. But SVG misses the concept of having documents with browsing contexts and without.

Ok but what precisely do you want on the HTML side? It's not clear that anything you're describing affects HTML. DOM, CSS, SVG, Fetch: sure. But why HTML? HTML already doesn't create a browsing context here, so the SVG spec can just describe how to render itself or fetch further resources when it is being presented in the absence of a browsing context, right?

I thought we updated HTML to only do style resolution (or whatever it's called) if there's a browsing context. But e.g. <html:style> would still work in such an SVG-as-image resource.

Ah, well, if you want certain things in HTML to work in certain cases where they don't work currently, please list what those should be and under what conditions they should work, and I'll be happy to update the spec.

Dirk, could you get the SVG WG to do that? SVG-as-image is the only case where this would apply.

https://github.com/whatwg/html/issues/702