| Newsgroups |
gmane.mail.ifile.general |
| Message-ID |
<[email protected]> |
[email protected] writes:
> My theory for the daemonized version went like this:
> - daemon locks the DB file .idata with lockf. (Is there a reason that's
> non-reliable that I should use an independent lock file instead?)
The lockf() call has historically been dubious over NFS and possibly
other shared filesystems. I wouldn't worry about it too much, as
nobody seems to have screamed about the current total lack of
inter-machine locking.
> - daemon creates a unix-domain-socket .idata.socket
I was about to say that you could use the existence of this socket
itself as a locking mechanism, but putting sockets in shared
filesystems is also dubious - other machines can't use the socket, and
some filesystems, like AFS, simply won't let you put sockets there.
I'd create a file .idata.lock and put the socket somewhere in /tmp.
> All that said, I think I'm out of cycles for this project for awhile.
Understood.
- Nathan