Should graphical-output challenges accept straight up image files by default?

Question

I've often seen some questions tagged graphical-output and kolmogorov-complexity that have some answer being image files, and they are often disagreements in the comments.

examples:

Swiss flag (PNG) Swiss flag (GIF) Iceland flag Carpet an airport Minecraft tool Super Mario RGB colour grid

Here's what comment section looks like (taken from the first example Swiss flag (PNG)):

Note:

This question is not realy about SVG; while they are some disagreements, it seems like it's more accepted than bitmap files. This is probably due to the fact that it isn't binary, but XML, which is closer to code than a binary bitmap image is. So you can talk about it but this is not the main focus of the question.

I gathered the arguments for and against it from the comments and some of my research:

For	Against
kolmogorov-complexity's description mentions a complex string, which shortest representation might be to print it literally. If a literal string is considered the "worst case solution" to a string based code-golf, a literal image file could be considered a "worst case solution" to an image based code-golf. - Sunday, in Carpet an airport	PNG isn't a programming language, this isn't "drawing" anything, it's just the pre-drawn output file handed out on a platter. That makes a profoundly un-interesting submission. - Caleb, in Swiss flag (PNG)
Answering in non-programming languages is allowed on Do submissions have to be answered with a programming language?	It could be compared to a compiled code (binary executable) which users seems to agree on the fact that they shouldn't be valid answers
It's kind of interesting that a generic commonly used image format like PNG can come close in size to the solutions using code tailor-made to this particular image - Anonymous, in Swiss flag (PNG)	This could give an unfair advantage, especially on pictures with a small colo(u)r palette. While I haven't seen any submission that wins, most of them are pretty close to first place

Answer 1

The thing is that not all image formats are created equal. Some are just straightforward uncompressed streams of pixels, and there is basically nothing you can do to golf those down. However if your format uses a compression algorithm, for example PNG, it is possible to hand tailor your input to exploit the compression algorithm. Even further down the line are formats like SVG (mentioned in the question), which encode vector graphics rather than a raster image. SVG golf is extremely non-trivial, you have to work with geometric reasoning and writing an SVG is much more like writing in a programming language than an image.

However there's a big difference between what is possible and what is done. Just because you can in theory write an interesting PNG answer doesn't mean that every PNG answer is interesting. I certainly agree that just uploading a plain, un-tampered, PNG as your submission is boring. It doesn't require any effort at all. It's not clever or fun.

There's not really an easy way to tease apart the interesting answers from the uninteresting ones. Or the interesting formats from the uninteresting ones. You say this question is not about SVG, but why? it is an image format. It seems you agree that SVG is more interesting, but who is going to make that call for every format? We already have an issue with unnecessary meta-cruft and rules.

It's just subjective.

But luckily we have a tool for that! We can vote on answers. If you think an answer is boring you can downvote it. If you think it's really interesting or shows high effort you can upvote it.

So I don't think there is any good reason to ban these. It doesn't actually achieve anything. Boring answers happen all the time no matter how many rules we have. But banning them blocks potentially interesting answers.