"Guy Lateur" <[EMAIL PROTECTED]> writes:
> To be honest, I don't really understand what it means to have the same file
> open for writing by several processes. You don't want to modify data which
> is already being modified by someone else, do you? I mean, how do you
> determine what changes to apply first, and to what version? Or is the file
> just constantly being overwritten on a first-come-first-served basis?
Unix believes (well, it used to...) that the programmer should be
given all the rope they feel they need. If that means they get enough
rope to hang themselves, so be it.
That means that when two processes write to a file, what they write
will be written to the file in the order the writes occur. So a later
process can well overwrite what an earlier process wrote. On the other
hand, it doesn't have to do that.
If the file being written has a binary record structure of some kind,
then it's perfectly reasonable for two processes to update different
records in the file "at the same time". utmp and lastlog are standard
Unix accounting files where this happens every time someone logs
in. Requiring each user that logged in to wait until all the previous
login processes had finished and closed the file would unnecessarily
delay a user logging in to a multi-user system. I've done this kind of
thing with dbm files, but later recanted and used a real database.
<mike
--
Mike Meyer <[EMAIL PROTECTED]> http://www.mired.org/home/mwm/
Independent WWW/Perforce/FreeBSD/Unix consultant, email for more information.
--
http://mail.python.org/mailman/listinfo/python-list
