| 摘要 |
A distributed (e.g., client/server) computing environment is described which implements protocol methodology simplifying the use of objects in distributed applications. In particular, the methodology facilitates streaming to clients executable objects (e.g., Java objects) stored and managed remotely (e.g., objects stored and managed in relational databases), so that the objects may be executed or otherwise manipulated locally at the clients. The methodology is implemented by extending an existing streaming protocol to support a "chunked" datatype; within a data stream, the system can have individual data items which are themselves streams of indeterminate length. This streaming data type is an undifferentiated data type or simply a "BLOB" (i.e., binary large object). Using the BLOB extension, the system provides a set of BLOB subtypes which take advantage of existing object streaming mechanism (e.g., Java streaming) but convey additional information in the form of self-describing metadata. The metadata contains all necessary information on the BLOB data for clients and servers to narrow the BLOB data itself to the appropriate subtype and extract the semantically correct values from it. In this manner, the system may employ the streaming protocol to receive requests and input objects and, in response thereof, generate and send output objects and output result sets with rows and columns that may include objects in them, or send back output stored procedure parameters that may be objects.
|
