I know of at least one person who is doing lots of science / scientific projects/experiments for themselves and does not have access via uni libraries etc. but needs to consult scientific literature (as in, academic papers) fairly often. (Granted, the case does not fall within the aforementioned 'theor.phys. / mathematics' field.) They use http://reddit.com/r/scholar to ask for papers to be uploaded (it's a very nice and useful subreddit for those folk, I try to fill out some requests now and then).
Of course, 'knowing one person' is anecdotal / talk about bad representative samples, but I can attest to the fact that those people exist. (I don't know the details why that person in question cannot register under a public library; public libraries may not have subscriptions to their journals of interest perhaps.)
In any regard, there are cases where old theories / research were dug up by contemporary researchers who realized that those original models were useful for their modern research etc. [citation needed, could dig something up, but googling is probably more effective than asking in this case.]
But he probably does it because it's just easier than getting a subscription to a university library. Proving my point exactly - how hard is it really to get access to academic papers? Most authors can't even find people that want to read their stuff. This whole 'I can't access research' is really 'there is no website where the 0.1% of people who don't have main stream and side stream access to journals can download them from'. Hardly the big threat to society it is sometimes made out to be.
Let me state again that I too would sometimes prefer one huge easily accessible database with all articles ever published, along with cites and H-factors and objective impact factor rankings and a bookmarking/personal library feature and maybe a pony too. Then again, the magnitude of that problem is miniscule compared to other problems I have I'd much rather see solved, or spend my time on. And my experience says most of the people in my work field feel the same.
I generally agree. At the same time, it's good to have tools for such edge cases - there are some papers in niche journals etc. which play a part in science but may not be easily attainable (again, because e.g. public libraries do not pay for subscriptions for those journals and whatnot.) But I agree with the gist of what you're saying (trying to come up with a universal user friendly system might be an overkill); but it is good to have these frustrations and debate in the open, and again, I really like that there is such a place as /r/scholar.
Of course, 'knowing one person' is anecdotal / talk about bad representative samples, but I can attest to the fact that those people exist. (I don't know the details why that person in question cannot register under a public library; public libraries may not have subscriptions to their journals of interest perhaps.)
In any regard, there are cases where old theories / research were dug up by contemporary researchers who realized that those original models were useful for their modern research etc. [citation needed, could dig something up, but googling is probably more effective than asking in this case.]