It just so happens that the field I was working in (astrophysics) is one where many papers are published on arXiv, but my supervisor always got frustrated when I used the arXiv copy of a paper (sometimes the final edit of a paper isn't uploaded to arXiv so you might be citing something which was removed after the journal review). And for papers not on arXiv, you can sometimes manage to the paper on the author's website but that's hardly a workable system (and those are usually preprints and not the final draft). And yes, you can get papers by emailing the author if you really need a copy but that limits your ability to search through many papers as broad background reading (it seems a waste of time for M researchers to ask N authors for a copy of a paper, only to read the abstract and results).
Academic papers (which are publicly funded through grants and often produced by public institutions) should be free to the public. Those papers have already been paid for by the public.
EDIT: ISO not being available was legitimately a real blow -- I was working on a personal project that required reading the DataMatrix standard and I couldn't get access to it. Instead I had to try to piece together the protocol from the Wikipedia description and just gave up because I couldn't get it working.
I work in industry in a position where having access to papers is very helpful, some would say essential.
I have access to IEEE and ACM DLs. I don’t, however, have access to myriad papers from other organizations like Elsevier or Springer Verlag. These are often cited by papers on IEEE/ACM DLs. Some are from 20-40 years ago and still very relevant and interesting.
I have yet to find a way to get access to these in a predictable manner (yes, sometimes you can find preprints or old PostScript versions on various authors academic websites, but not generally).
Often the contents are from government sponsored research from the US government and possibly other governments (if only through their support of Universities in their country).
Edit: Just to add insult to injury I published 2 papers while in my bachelor's and I could not access them because they were behind a paywall by Elsevier.
Before 1990 the activity of a scientific journal was somehow more complicated and costly which somehow justified the subscription cost. With the current Internet infrastructure and tools, paywalls feel more and more like symbols of legacy monopolistic economy rent.
You should at least glance at the table of contents of https://en.m.wikipedia.org/wiki/Open_science and read those short sections "Publicly funded science will be publicly available" and "Disadvantages".