As per this video, browsers will begin laying out images based on height and width attributes so they don't "jump" on load.

Of course, with user-generated content, we can't always know the height and width of an image, but we could.

We do various forms of image manipulation in the markdown_parser, i.e.

    html = prefix_all_images(html)
    html = wrap_all_images_in_links(html)

And in that process we set some attributes such as img["loading"] = "lazy". In that area we could also set img["height"] and img["width"].

In terms of "how" to do this, the library FastImage looks like a good one to me, but I didn't do an extensive search. I'd think we'd want to create a method which Rails caches the output of the image we fetch, so we don't need to make this network fetch every time we hit "preview" or re-save an article. FastImage also lets us set a timeout, which we should definitely make use of because this is not critical behavior we'd want to cause a major slowdown with.

Anyway, if we can make this work without making things meaningfully more fragile or slow, I think we should do this. Having awareness of aspect ratios seems generally useful.