Classification schemes are an important issue in the collective use of large document collections. We have investigated the classification of technical documentations in two engineering domains: a steel mill and a sewerage plant company. In both cases we found a coexistence of different classification schemes and problems resulting from distributed local archives. In supporting human actors to maintain different classifications schemes while working on a common archive, we developed the concept of context grabbing. It allows assigning context information efficiently in the form of metadata. Based on a document management system, a tool kit for context grabbing was developed. Its evaluation in a sewerage service company allows us to comment on important aspects of understanding the role of classifications in collaborative work.