I'm afraid you'd need to explain a lot more, to make people install this. Like: how does your rating work and how would it be better than the votes we're already seeing?

Be prepared for a link to I LOVE JPEG!! JPEG! JPEG! when posting screen captures then ;-)

I wonder if you really need to know the image size? If someone enters a width less than 320, then you can use the m thumbnail? But even then: it may still NOT be worth the effort, as you cannot do the same when someone enters a value for the height (unless, indeed, you would know the original dimensions, and hence the aspect ratio, to do some math). Another aside: given the fixed post width of about 600 pixels on the regular Stack Exchange sites, and much less on the mobile theme, I suddenly think we all could always use the l image? Too bad that's making PNGs into JPEGs then.

And lunboks wrote that the resized images are always JPEG, which seems to be true. So, maybe change the extension too (just hardcode 'l.jpg' rather than 'l.$1')? It works with the original extension too, but it's just not 100.00% right. ;-) (In one test I saw some binary PNG data, not JPEG data, but I cannot reproduce that. Things might be different if there's no need to rescale?)

lunboks wrote that l and m have no effect if the image is already less than 640 and 320 pixels wide, respectively. Would it make sense to add an l or m if someone adds a custom value for width? (I'm afraid it will only complicate things; it's just that using the original image with <img width=...> does not lower the bandwidth for those who're not interested in the actual image. But maybe you already thought about that.)

Given the CC-attribution-required, I'd like the author's name to be on the print too. Other than that: nice!

When using The Printliminator bookmarklet, one can quickly remove certain parts of a page. To use that effectively: would it be feasible to add a parent <div> per answer, and for the comments?

No worries, if a 500 error ever shows again, the search for the binary data will point folks to this very question! ;-) (Well, at least the Google search does!)

But I guess it's hard to make error pages return the correct encoding headers too? (Just like the 200 OK response does.)

(I got a blank page yesterday, @George, but now I get a virus warning. It's telling me "This IP address has been seen posting comment or blog spam", while earlier I guessed it was caused by some blocking I do myself, or by using a non-compatible browser or OS. I don't think I like that "This website is participating in a project to stop attacks and educate visitors with infected computers about how they can clean up their machines" if it is indeed throwing false positives. Right now, the scary cat bothers me more! ;-) But I'll investigate later...)

True, @mmyers, but I think the original images should be preserved. Like I wrote: the revision history will show the author's original image if someone else added freehand circles at a later time. But even then: not too nice if a post shows no image at all without going into the revision history?

(I'm getting an "Access Restricted" for that image, and the entire placekitten.com website. The page source has an odd iframe, referring to http://anti-virus.cloudflare.com. That company apparently also protects against hotlinking, which seems to be odd for this very service. I can still get there using Web Sniffer and then get totally different HTML. Maybe my IP has been blocked due to the above hotlinking? Odd!)