This month's most interesting messages:

All from the SETI@home team members at Berkeley. Non-official statements but interesting anyway.

Hiram Clawson - 2000.03.03 in a.s.s., client problems:

As all ya'all have noticed, something unusual happened to the stats on March 1st. We are still not sure what it is. It seems to be a problem at the server receiving end. It seems to be related to the version 1.x expiration of the clients:
sparc-sun-solaris2.6, alpha-dec-osf3.0, i386-sco-sysv5.unixware7,
i386-pc-sco3.2v5.0.x, i386-pc-sysv4.2uw2.1,
i686-pc-linux-gnu-gnulibc2.1 and i386-sequent-sysv4

Which happened this past weekend. The WU return rate for all of these clients mysteriously jumped on March 1st. It is merely a problem of properly counting results. The science has not been jeopardized, and these clients are not actually returning more WUs than is normal.

We are looking into the problem to see what happened.

Eric J. Korpela - 2000.03.07 in s.a.s./a.s.s., about the v2.03 bugfix:

The [ >8 ] slower write frequency is part of the fix. On fast machines it was easy for the result file and the state file to get out of synchronization on disk due to the caching. A crash or imperfect exit (another bug that was fixed) would result in either the work unit restarting entirely, or backing up several steps.

Hiram Clawson - 2000.03.12 in s.a.s./a.s.s., the subject of running S@H on computers for which you are not authorized:

This week's Science News 04 Mar 2000, vol 157, no 10 has a story about GIMPS, S@H and other distributed computation projects. It has a story about one user of GIMPS who was arrested for using company computers for the computations.

http://www.sciencenews.org/20000304/bob1.asp

Please do not use computers for which you have not been authorized!

Hiram Clawson - 2000.03.12 in s.a.s./a.s.s., the porting process:

Here's what the managing of the porters entails.

There are about 30 porters of the client for the UNIX ports. A port begins by sending out email to the lot advising them of the availability of new code.

I do five of the ports myself. The process is automated for picking up the code and building it on the five platforms. Simply fire off a small script. Then each of those five ports has to run through a test work unit. Login to each machine, run the client, make sure it looks sane, get it started on the test work unit. When that finishes, a bit of regression testing to verify that the client is OK on a couple of known problem WUs. Thanks to all of you for letting us know about particulare WUs that crash the client, etc.

In about two days, two or three of the other porters have done the same thing, their test results have been returned, and their client binaries. They advise me by email. I run a script that summarizes all returned results into an html page where I can easily see if everything is passing OK. If that is all right, I package the binaries into the tar image and place them on the ftp directories at Berkeley. And set the data base server to accept results from those clients. I email those porters to advise them to check their tar file to ensure I packaged the correct items. They respond with an OK. A couple of days goes by, the tar files have mirrored to ftp.cdrom.com, I can now remake the unix.html download page.

Over the next two weeks or so, perhaps two or three other porters will respond with their completed work. Same process for each of those.

Now the work begins. Examine the listing of completed clients and compare it with the clients to be done. Create a list of porter email addresses for ports not complete. Email that group to query status. In return, receive some very creative explanations for delays and completed clients begin to trickle in. Repeat this process seemingly forever. Identify the high use clients and hit on the porters that haven't done those yet. Discover some porters have disappeared, examine email archives to sift the hundreds of offers from all ya'all to help with the porting. Establish contact with new porters, bring them up to speed on the process. Discover that some porters are not actually programmers. Help them through the difficulties of solving simple compile problems on their various operating systems.

Discover that a particular client is broken. Redo the process with the responsible porter. Exchanging emails for all of these transactions with the porters.

While all that is going on, make it look like I'm still doing my job at work, getting to aikido practice, gardening, spending time with family and friends and continue to wonder why I like this volunteer job so much. Isn't this a great project!

Matt Lebofsky - 2000.03.20 in s.a.s., regarding all those funky screen shots:

It has nothing to do with location in the sky as much as when it was recorded. Some tapes might have short periods of time on them after other groups at Arecibo started transmitting data and before our data recorder got disabled. What you are seeing is probably interference from right there at the observatory, or something quite similar.

Hiram Clawson - 2000.03.21 in a.s.s., about transferring WUs between computers:

The client binary and its associated *.sah files form a complete unit that, once processing has begun, can not be disturbed. This is a function of the new security features of v2.0 to prevent tampering with the files or the client binary.

If you want to transfer partially completed WUs between computers, you must have identical client binaries in both places.

You can download WUs with one binary, transfer that WU to a different client, allow it to process the WU to completion, transfer the results back to the first client and it will upload the results. In this case, the clients can be different.

Hiram Clawson - 2000.03.28 in s.a.s., same old questions about the WINNT CLI 2.4:

The v2.4 clients do NOT output the %done messages. Use the -verbose option if you want to see those messages.

TO ALL USERS: PLEASE NOTE THE MESSAGES ON THE DOWNLOAD PAGE:
http://setiathome.ssl.berkeley.edu/unix.html
and read the FAQ:
http://setiathome.ssl.berkeley.edu/faq.html

to answer probably %98 of all queries that go by here every day.