I'm not too unhappy with Google Scholar lately. It seems to have nearly complete coverage of anything that can be found in an online paper archive (both institutional archives and journals' own archives), and a better search interface than those archives do. Via extracting references it even indexes a large amount of older material that's not yet digitized; of course it can't link you PDFs in that case, but it still provides the citation and can return offline papers by title/author/date in search results.
Google Scholar has a reasonable interface, but I find its actual content to be unreliable at best. Downloading references in e.g. BibTeX format, I almost always find significant errors, sometimes in relatively innocuous fields (different formats of a conference name, etc.) but other times in more important fields (authors' names).
It's a great tool, and I use it all the time, but it's certainly no replacement to a properly-managed bibliographic database.
Yeah, I use it purely for discovery, not for bibliographic management. I usually just type in BibTeX data myself rather than importing it from anywhere. Often PDFs have the needed info (authors/title/venue/pages/year) inside the document itself, and if not, you can usually search for the article's title and find the info on the publisher's site. In CS, DBLP also has reasonably good curated bibliographic data, but lower coverage and no full-text search.
I'm not too unhappy with Google Scholar lately. It seems to have nearly complete coverage of anything that can be found in an online paper archive (both institutional archives and journals' own archives), and a better search interface than those archives do. Via extracting references it even indexes a large amount of older material that's not yet digitized; of course it can't link you PDFs in that case, but it still provides the citation and can return offline papers by title/author/date in search results.