Journals act as the Yahoo of scientific papers. That is they are an index of certain topical relevance and editorial quality. (The original idea of Yahoo was a vetted directory to web pages) I am not sure if papers scattered around random university servers would be easily found.
A case in point is the annual proceedings of the SIGGRAPH computer graphics conference. ACM sells a wonderfully color printed volume of these papers for nearly a hundred dollars. However an individual keeps a web index to the half of these papers posted on private laboratory websites. The index is free, but the quality of printing varies. This private index has already been vetted by SIGGRAPH for conference quality- the only accept about 1 in 15 submissions. Occasionally I poke around distinguished computer graphics labsvwesites. But their quality is variable. Sometimes the website is abandoned when the grad student care taker moves on.