[jcifs] Mike or Chris

[Dan Dumont]

Without trying to connect to a host, how will I know if it can accept
connections?  With this line uncommented, the only thing I saw was that the
maxmpxcount would be reduced to about 5, which would mean that only 5
computers would be scanned at a time, all of which fall victim to the the
1.5 minute timeout.

I can send you the code that I have, provided you can get a long listing of
ip addresses.   You can see the performance difference, I think it may
surprise you.

-----Original Message-----
From: Allen, Michael B (RSCH) [mailto:Michael_B_Allen at ml.com] 
Sent: Tuesday, April 29, 2003 6:44 PM
To: 'Dan Dumont'; jcifs at lists.samba.org
Subject: RE: [jcifs] Mike or Chris

Dan,

If you comment out that line you are basically letting jCIFS create
unlimited transports. This will permit jCIFS to very quickly make GREAT
progress in creating hundreds of useless connections to
hosts that do not exist and will basically result in nothing but wasting
GREAT amounts of resources. I doubt overall performance increased. It would
also be possible (although highly unlikely in your
context) to create more than the negotiated conurrent requests acceptable to
a server.

I have tried to explain as best I can why you're having this "problem" which
also explains why your perl script runs more quickly. I would suggest you
read up on Transmission Control Protocol (TCP).
It's what Java uses to create stream based sockets like those used by jCIFS.
Then go back and review my original responses to your questions
(particularly the part about the lack of asyncronous IO
facilities in Java).

Mike

> -----Original Message-----
> From:	Dan Dumont [SMTP:Dan at canofsleep.com]