Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uunet!dptechno!dave From: dave@dptechno.UUCP (Dave Lee) Newsgroups: comp.databases Subject: Re: db_VISTA III -vs- Relational DBMS' Message-ID: <554@dptechno.UUCP> Date: 9 Aug 90 20:27:16 GMT References: <15660@s.ms.uky.edu> <1990Aug8.144318.10203@dg-rtp.dg.com> <2660@lamont.ldgo.columbia.edu> Reply-To: dave@dptechno.UUCP (Dave Lee) Organization: D.P. Technology Corp. Camarillo California Lines: 45 In article <2660@lamont.ldgo.columbia.edu> paulf@lamont.ldgo.columbia.edu (paul friberg) writes: >Another addition to the list of problems with db_vista III is its >lack of networking (i.e., client/server communications). This is >fine if you are operating in a homogeneous network where you can >NFS mount your data disk on every host from which you wish to access >the data base. For a heterogeneous network, however, you are basically >screwed. > >Can anyone offer any insight into db_vista as to how it does >its IPC? I use db_vista III under unix and have developed a product that uses it in a multi-user,multi-computer network. DB_vista has support for multi-computer networks under DOS, using LAN manager system locking. Under unix, there is direct support for multi-USER, single CPU systems. Data is stored in binary form in the host`s native format. IPC is through the OS-specific system calls (message queues on SYSV, sockets on BSD). There is support for a system independant locking stratagy that relies solely on lock files in a common directory. This technique is not recomended due to high file access overhead. This MAY work on multi-computer networks using NFS, but I havent tried it. However, all CPU's must share the same basic data representation. The system I constructed used a server host running a db_vista back end for every active user. A front end on the client cpu communicates with the back end on the host via a RPC interface. All data transmited to/from the backend is in ASCII and encoded/decoded at each end. Also high level commands are implemented such as update,create,query ... I find this scheme quite acceptable, although it did require a good deal of my coding to implement. -- Dave Lee uunet!dptechno!dave