[PROXYPER] Personal Proxy 304 Q&A
jlawson at bovine.net
Mon Feb 15 01:32:11 EST 1999
Although it would be possible to implement locking in various other
alternate mechanisms, such as using a separate accompanying file, it would
not significantly improve things. A method is provided for unlocking
files via the -unlock switch.
Separate lock files can also cause false results, such as when the
specified file you are attempting to "lock" is actually a symbolic link,
since it is questionable whether the lockfile needs to be placed in the
directory where the symblic link is, or whether the symblic link needs to
be resolved, and the lockfile placed in the actual file directory.
Symblic link resolution is a more involved process, but without the
resolution (and thus just placing the lockfile within the directory of the
link), it would be possible for the lockfiles to be completely ineffective
if two pproxies were being run from separate directories, but accessing
the same buffers through different symblic links leading to the same file.
Assuming that if a specific pid is not currently present then the proxy is
not running is not necessarily safe, particularly in environments where
the directory structure is cross mounted (or fileshared) among multiple
Regardless, the current system of locking works reasonably well and is in
fact the same manner in which the client currently uses, so I'm inclined
to leave it as is.
Jeff Lawson http://www.cs.hmc.edu/~jlawson/ http://www.bovine.net/
Jeffrey_Lawson at hmc.edu jlawson at bovine.net bovine at distributed.net
Programmer, Developer, Mascot, Founder of the largest computer on earth!
Don't waste those cycles! Put them to use! http://www.distributed.net/
On Sun, 14 Feb 1999 rc5 at xfiles.nildram.co.uk wrote:
> That would be a nice feature (at least for *nix) proxies. newsx does
> something similar. It checks to see if another process does, infact,
> exist, and if it does it signals it and it exits. If it doesn't it
> removes the stale lockfile.
> On Sun, 14 Feb 1999 root at brain.acmelabs.com wrote:
> > > The "problem" of buffers getting "locked" with the 300+ proxies and the
> > > new need to unlock them should not be considered a weakness. It is
> > > designed as a protection mechanism to attempt to eliminate conditions in
> > > which users accidentally start up two instances of the proxy, and have
> > > them both transmit the same completed blocks upstream, which was an
> > > occasional problem for us, which resulted in duplicated done blocks being
> > > reported, and incoming new blocks getting lost.
> > Why not use a seperate lockfile? one with a process number inside.
> > Maybe the proxy should actually check if there is another proxy running
> > instead of assuming and refusing to run. Hmm.. now that would be nice.
To unsubscribe, send 'unsubscribe proxyper' to majordomo at lists.distributed.net
More information about the proxyper