The Citeseer Scientific Digital Library contains downloads for many excellent papers: http://citeseer.nj.nec.com/
ACM provides some free papers, but IIRC to view many of them you must be a member: http://www.acm.org/
From RFCs (for Internet protocols) to USENET newsgroup FAQs: http://www.faqs.org/
For RFCs specifically: http://www.rfc-editor.org/
ARPANET papers, which are useful if you want to know about computing history (especially the Internet): http://www.archive.org/texts/arpanet.php
USENIX has many research papers, and I believe most of them are free to view: http://www.usenix.org/
arXiv (http://www.arxiv.org/ ), formerly known as xxx.lanl.gov, is a huge preprint archive. It started as a physics preprint archive, but it now covers also mathematics and computer science.
PhysNet (http://www.physnet.net ), Physics' departments and documents
HAL (http://hal.ccsd.cnrs.fr ) is a french server for depositing scientific papers, it covers physics and other disciplines available from arXiv and has a write through to arXiv for those papers. But it is open for any scientific discipline.
citeulike (http://www.citeulike.org ), A kind of social bookmarking for scientific papers.
ro If you know the title of the paper you're looking for, then enter it into google with '+ pdf'. If it's at all popular, it's out there.
VL It will also be on the internet if the author is computer savvy enough and interested in 'publishing' it, and yes I use Google. If you don't know the name of the paper, go to a university library site; they tend to have subscriptions to abstract databases and such that are free to search.
DKF: Google Scholar (http://scholar.google.com ) is pretty good.