I'm starting an interesting project for my company. We have a large, complex proprietary legacy database with a considerable suite of backend software that maintains it. The client applications are NT-based and access these flat files through Windows Networking or Netware. Our mission is to turn it into client/server software, supporting multiple tiers of servers over TCP/IP. Also we'd like to have deliverables as we go along, rather than making it one huge monolithic project. What I decided to do is to create a server that encapsulates the legacy database (which in the short term will not change) and emulates a simplified PostgreSQL server. This way we get immediate SQL connectivity for new client applications, and the ODBC/JDBC drivers are already done. Initially this is happening under NT but I have fantasies of porting to Linux. I've already got quite a bit done, and have code acting as a "passthrough" to a real PG server, and displaying a line of tracing information for each message going back and forth. It's fully multithreaded. Next we'll start plugging in the legacy database and parsing the requests and delivering appropriate responses. Of course this will never be a general-purpose SQL server, as it only needs to deal with a particular kind of database. We'll eventually design some kind of protocol so that the server can replicate its data to another server at a lower tier. I don't know what to call this kind of approach... it makes a lot of sense to me, but has anyone done this before? All comments are welcome. Thanks! -- Rod
