Wednesday, June 16, 2004

WinFS, advertised as a way to make searching work by making the file system be a relational database, ignores the fact that the real way to make searching work is by making searching work. Don't make me type metadata for all my files that I can search using a query language. Just do me a favor and search the damned hard drive, quickly, for the string I typed, using full-text indexes and other technologies that were boring in 1973 [Joel on Software] Proponents of rich metadata believe that if we only had the right ontology, and users were a little bit less lazy, search would be ever so much better. The ignore the economics. Creating metadata is a significant up-front cost, which in most cases is not recouped because most items are saved just in case they are needed, although in fact they never are. They also ignore the constant semantic shifts that undermine old metadata as knowledge, forms of expression, processes, and code change. Except for very specialized, high-investment data items that justify carefully hand-crafted metadata, the only metadata that matters is automatically generated metadata. Search engines already generate metadata when they recognize automatically types of information with particular semantics, such as addresses or product names. This is useful even if not perfect, and the range of types and accuracy of automatically tagged items are improving all the time. I don't think that we knew in 1973 all that is needed for efficient and accurate ranked search that takes into consideration types of information and their relationships. But better models of user information needs and better search algorithms are much more important than a new file system. 9:46:13 PM