Re: Raw partitions / cooked files

In article <1993Jul8.145915.19497@informix.com> davek@newjersey.informix.com (Dave Kosenko) writes: ............................... not to mention those of integrity (a successful >return from a write() does not guarantee that the data is on disk when a cooked >file is used; the data could be sitting in a Unix buffer, which can result in >data loss if a crash occurs before that buffer is flushed to disk). Not true. LOGICAL LOG writes are synced writes if the log resides on a block special device (filesystem file). This is all the matters. If there happen to be 400 modified buffers in Online's Cache at the moment of failure, what's the difference? Data Consistency is not an issue of FS vs Character Special files...yeah I know, the indirection of tbpgcl flushing without sync to a filesystem file may catch non-synced Online buffers in the Unix cache at the time of failure, but it doesn't matter. Fast-recovery goes back the the last successfull checkpoint before the failure and reapplies ALL COMMITTED transactions AND rollsback those that didn't make it to commit. ________________ Kevin A. Closson, Senior Technical Support Engineer voice: (503)578-3072 World Wide Technical Support, Database Group FAX: (503)578-5453 Sequent Computer Systems 15450 SW Koll Pkwy (MS WIL1-502) UUCP: ..uunet!sequent!kevinc Beaverton, Oregon 97006-6063 internet: kevinc@sequent.com