We conducted a study on how academic researchers manage multiple documents acquired from the web for later retrieval. We interviewed 11 participants and identified their strategies when trying to re-find specific documents. We found that they often prefer web-based search for re-finding documents, despite knowing that the document of interest is stored on their computers. We argue that Web search engines can act as an extension of the desktop information space. We found that users choose keyword-based search not only when the document's location is unknown but also when the retrieval cost is very low: they do not bother about properly storing files because most files are easily found again with a web- based search engine. We close by discussing the implication of these findings for the design of future document management tools.