Dessy: A Desktop Search Framework for Mobile Devices


In the near future, mobile devices are expected to have a storage 
capacity comparable to today's desktop machines. As the amount of 
information grows, conventional search tools become less effective. To 
meet these challenges, we turn to desktop search. I present Dessy, a 
desktop search framework for small devices and desktop computers. It 
allows a user to find files by their content, metadata, and context 
information, provides an interface for locating files for both users and 
applications, and allows finding files with just their metadata 
available. Custom metadata “tags” as found in other desktop search 
software can be added to files with Dessy. The framework supports 
separate synchronization of file metadata and data. Dessy uses the Syxaw 
file synchronizer. Synchronization of select file groups such as search 
results, subdirectory trees, or the whole file system is possible. Dessy 
is unique among desktop search software in that it clearly separates 
extraction, exchange, and querying of search metadata. In particular, 
metadata extraction may be performed on one device and querying on 
another, with only search metadata being exchanged between the devices. 
The metadata is exchanged in a simple plaintext format well suited for 
network protocols. Dessy emphasizes extensibility: adding new indexed 
file types, metadata fields, and metadata storage methods is easy. This 
enables a small device to use simple metadata storage while exchanging 
metadata in a common format with a  desktop machine that uses a 
full-fledged database to store metadata. Finding files is done with 
virtual directories, that look like a file path and represent queries to 
Dessy. Reading a virtual directory shows the files found by the queries 
represented by the path of the virtual directory. The virtual 
directories can be browsed by regular file managers through an NFS 
interface on desktop machines, and using an API on small devices that do 
not support NFS. To better match queries to metadata words, all indexed 
words and queries are stemmed. An implementation of the framework and a 
working prototype system are in development at the Helsinki Institute of 
Information Technology. The presentation includes a demonstration of the 
prototype.