From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On Behalf Of Kevin Zollinger Sent: Wednesday, October 02, 2002 7:35 PM To: [EMAIL PROTECTED] Subject: Re: Serialization And Unix File System Semantics Question
----- Original Message ----- From: "David T. Ashley" <[EMAIL PROTECTED]> > > Hmmm ... from the URL you cited, I obtained this: > > <BEGIN> > write( str) : Write a string to the file. There is no return value. Due to > buffering, the string may not actually show up in the file until the flush() > or close() method is called. > <END> > > Statements like that do not inspire confidence. > > I'll research it to the end. Kevin wrote: That is standard practice, and used to improve performance. Keep in mind though that there is but one buffer for each file device, and that while the particular bits in the buffer may not be written to the physical media immediatly they will be written sequentially. Go ahead and do you research, but you're not going to find anything.... -------------- Kevin, If the buffer is on a per-file basis, no problem. That is expected. If you are saying that all processes with the file open use the same buffer, then there is not a problem. But if the buffer is per process ... The statement "may not actually show up in the file" is a cause of concern. By my definition, info that is cached but not yet written to disk is still "in the file" as long as any process that opens the file gets the cache rather than the disk contents. Dave. _____________________________________________ tmda-users mailing list ([EMAIL PROTECTED]) http://tmda.net/lists/listinfo/tmda-users
