Unfortunately, that's not the case in pragmatic reality.
When I say #ff0000, do I mean "sRGB red", as a TV would display in pure red phosphors back in the 1970s, or do I mean "1000K" / 700nm red (daylight is 6500K), which can only be reproduced by specialized displays?
Most people from the time before Apple made Display P3 wide color displays standard in all their products — so, most HN commenters — believe that #ff0000 just means "only red, no green or blue" — but all web browsers currently are left to invent their own answer for what #ff0000 means, and they do not all invent the same answer. Yet.
So it is with images.
In the beginning times, people made sixteen-color images for NTSC monitors, and then other people learned you could display synthetic colors that don't exist by mangling the image bitstream to the monitor, and those sixteen colors were hard-coded but varied by like +/-5% per device due to acceptable manufacturing tolerances, so they were internally consistent anyways.
And so we end up, today, trying to solve color profiles for file formats that specify colors as RGB hex values — which are, by definition, restricted to sRGB and thus wildly insufficient. But if we plan too far, we get file formats that are totally ridiculous - TIFF comes to mind - that can represent anything under the sun, at the cost of having a 1:1 mapping of "pixels" to "bytes" or worse.
You may also find the history of the EXR format relevant here, as it supports a functionally infinite amount of dynamic range in images, and there are definitely good niche use cases for it — but pragmatically, it's ridiculously overkill for normal everyday purposes. You could argue that everyone should use EXR in Lab, but then someone would suggest Oklab (since it makes better perceptual compromises) or HSLuv (what a great name), or you could try storing raw CIE 1931 chromaticity coordinates, which has been deprecated by CIE 170-2 to address serious flaws in the 1931 system.
Tying this back to engineering, there are often very good reasons for designing a processor or a byte store to be RISC or CISC, big or little endian, and we simply cannot predict what will work most pragmatically for all cases universally ahead of time, any more than the metric system was invented early enough in time to prevent miles per hour from being a unit of measure.
