Berkeley CSUA MOTD:2004:January:02 Friday <Thursday, Saturday>
Berkeley CSUA MOTD
2004/1/2 [Computer/HW/Memory, Computer/HW/CPU] UID:11644 Activity:high
1/1     I got a IBM R40e recently with windows Xp.  I'm wondering what I should
        run (if anything) to stress the computer out.  I heard that if a
        laptop is going to break it's going to break in the first few months of
        use.  Should I worry about this?  I'd much rather ship this back in
        30 days to the seller rather than deal with IBM warranty.  Thanks.
        \_ setiathome
           \_ bah. recent games are still generally the things that work a
              computer the most.
              \_ In terms of actual stress testing, not really.
              \_ yeah goof point. seti only stresses the CPU.
        \_ Video editing and reencoding.
        \_ I had a Dell Inspiron that had an overheating GeForce GPU --
           only found out after playing Flashpoint on it for a day.
           It also took two or three exchanges with Dell before they figured
           it out.
        \_ M$ Fight Simulator in demo mode with highest randering options.
           That's how I stressed a PC I bought a few years ago.  Turned out
           that it hung after 15min or so and it was a CPU problem.
        \_ Prime 95 has a good stress mode, not for the video card though. It
           was able to detect faulty CPU operation when overclocked, even though
           stuff would appear to run ok.  You have to run the
           specific stress test. Some people run two instances.
        \_ is it just me or it seems like modern computers are less fault
           tolerant than ever?
           \_ It's just you. Personal computers are much more fault tolerant
              now than they used to be. Apparently you've never run an AppleII
              or a C64. People's expectations change and they forget about the
              recent past. It would have been ludicrous during the dawn of the
              PC era to think that a C64 could have uptimes of months. You were
              lucky if it didn't bugger out after a couple hours. You can't
              compare today's multipurpose PC to mainframes of the past because
              that's like comparing apples with oranges. The fact is that PCs
              have essentially replaced mainframes because they have become
              increasingly more fault tolerant to the point that they can be
              reasonably used in an enterprise market. That doesn't mean that
              they are as fault tolerant as a mainframe, but the fact is that
              you can keep a Linux/BSD x86 box running for months shows the
              vast improvement over time of the OS and the hardware archtiecture
              of PCs.
              \_ By "fault tolerant" was the poster asking about how often
                 something goes wrong, or how likely the machine keeps going
                 after something has gone wrong?
           \_ Not all modern computers are designed to be fault tolerant. If you
              have a plain Pentium IV and non-ECC RAM, you aren't getting the
              greatest reliability but probably enough for most people. With a
              Xeon and ECC RAM, you get better data integrity in both the CPU
              and from the RAM. Enterprise storage equipment has more. I guess
              data integrity isn't the same thing as the real fault tolerant
              stuff on servers like RAID and other redundancy, which is what
              you would need to tolerate disks, psus, cpus etc. going bad.
2004/1/2 [Politics/Domestic/California] UID:11645 Activity:nil
1/1     Voters Crossed the Line in Miami
2004/1/2-11 [Computer/SW/SpamAssassin] UID:11646 Activity:low 54%like:10317
1/2     SpamAssassin 2.61 installed (finally); bugs to mconst.
        \_ if you specified "| spamassassin" in your .procmailrc recipe,
           please change to "| spamc" instead -- it is lighter weight/faster.
        \_ now that sa does bayesian analysis, what are the pros of ifile?
           \_ It's the standard.
              \_ Who thinks that, other than you? Do you have URL or
                 anything suggesting that "it's the standard"?
              \_ like IDS, it is better to us spam catchers that are NOT
                 the standard.  -phuqm
        \_ Spam Assassin rocks...  Say goodbye to the Nigerian scam, toner,
           FREE VIAGRA, and the lot.  Awesome.
           \_ Thanks, joshk
        \_ 3 days of spam-free inbox and counting.  Thank you!
           \_ It's only a matter of time before those fuckers figure out how
              to get past the filters again.
        \_ I'm loving it!  It's great, Thanks.
           \_ Damnit. My brain immediately coughed up McDonald's. Yay America!
        \_ Thanks.
           \_ I did this. The next day there was lots of spam and no indication
              that spamassassin had been working. I also see nothing via ps
              to indicate a spamassassassin-related process running. I'm
              switching back to the old executable line because I know it
              \_ please don't do that:  spamc and spamd work together,
                 and they are much more efficient.  If everybody ran
                 "| spamassassin" like you do, soda would be really slow.
                  /usr/local/bin/spamc   man spam    ps aux | grep spamd
                  \_ If it worked for me as you have it here, I would use it.
                     It doesn't, so I don't.
              \_ Not to mention, spamc is forked by procmail only at the
                 receipt of new mail- it's pretty transient.   RTFM.
                 \_ Uh-huh, but spamd or some such should show up via
                    'ps -aux' regardless.
                    \_ try 'ps -auxww' you ingrate.
                      \_ How dare you post useful information! (thanks)
              \_ IFile still catches about 50% more of my spam than SA
                 \_ SA caught 137 spams for me with one false positive.
                    No spam got through for me. -ausman
              \_ I also switched from spamassassin to spamc and there was
                 no indication it worked. Maybe that's why it's so fast!
              \_ spac is a standalone program, no need to call perl
                 was that your problem? it bit me
              that spamassassin had been working. I'm switching back to the old
              executable line because I know it works.
              \_ Could you please send me mail, so I can try to figure
                 out what's going wrong?  --mconst
                 \_ I'll try switching to spamc again today. If it appears to
                    not be working when I check mail tomorrow, I'll mail you
                    then. Thanks.
              \_ okay, here's exactly what you do, put these in your
                 .procmailrc before your other recipes:
                | spamc

                * ^X-Spam-Status: Yes

                next if you have any spams that fall through, you sa-learn
                to teach SA to filter those.  See man sa-learn
        \_ two stupid questions:
           1. is there a better way to invoke procmail than through .forward?
           2. is there a better way to disable the report text with spamc than
              spamc -c ?
           \_ 1. you don't need .forward at all, just .procmailrc
              2. man Mail::SpamAssassin::Conf, or just put "report_safe 0" in
              \_ 1. what calls procmail then?
                 2. man Mail::SpamAssassin::Conf
                    -> No manual entry for Mail::SpamAssassin::Conf
        \_ I'm having a problem with sa-learn, it keeps telling me:
           soda [34] sa-learn --spam /tmp/uncaughtspam_20040109
           Learned from 0 message(s) (1 message(s) examined).
           However, I have over 250 spam messages in the file. -gsu
           \_ RTFM: "sa-learn --spam --mbox /path/to/mbox"
              \_ I RTFM and the man page must be out of date because it says:
       Use this tool to teach SpamAssassin about these samples,
       like so:

                   sa-learn --spam /path/to/spam/folder
                   sa-learn --ham /path/to/ham/folder
                 Thanks though, because --mbox works.
                \_ notice that the example says /path/to/spam/folder, NOT
                   /path/to/spam/mboxfile.  Without --mbox, it's expecting you
                   to point it to a folder containing each message as a
2004/1/2-4 [Uncategorized] UID:11647 Activity:nil
1/2     Your application, That movie, Wicked screensaver, Details.
        Over two thousand emails in one day. what was the name of that virus?
        Has it returned?
        \_ i also received about one thousand starting Dec 31, addressed from
           www-owner. is this also the case with you? -jwang
           \_ the ones I got were addressed to www@csua. which I am on.
2004/1/2-5 [Consumer/Camera, Computer/Companies/Google, Computer/Theory] UID:11648 Activity:nil
1/2     Computer science question for you PHDs out there.  What would be
        the feasibility of a program that, given a set of images that form
        a mosaic, produce the "most correct" composite image, based upon some
        definition of correctness that could be supplied in advance (color
        compatibility, smoothness of lines, etc)?  For instance, such a
        mosaic could be a 360 series of photos that form a panoramic
        photograph.  Such a program wouldn't necessarily need to be perfect,
        and any of amount of "hints" could be given to the program as well
        as the input images.  This might already exist, or it might be solving
        the halting problem - I don't know.  I'm asking the question for a
        materials science post-doc friend of mine that is working with
        crystal lattice images.  --lye
        \_ Does the camera rotate as it takes pictures?  Is there overlap
           between individual images?  At its most general, this problem
           involves object recognition and so is vision-hard.
           There are some papers on this problem, google for obvious things
           to find them.
        \_ such programs exist. if the camera undergoes pure rotation
           and no translation about the optical center, the problem is
           very easy to solve (assuming overlap between the views). other-
           wise it's harder and you have to rely on some kind of
           approximation because you need to know the 3D geometry of the
           scene. other things that help: if you know the exact motion of
           the camera, the problem is easy again. -ali
        \_ There is a ton of literature on this kind of problem, which is
           known as "registration." A standard approach is to define some
           kind of error function (distance between edges, or distance
           between overlapping pixels in color space, etc) and try to minimize
           it over the space of transformations. An algorithm that works well
           for a lot of problems of this type in the pairwise case is called
           "iterative closest point," due to Besl and McKay. If you don't
           know an approximate solution to start with, it is a lot
           more difficult. -lewis
           \_ homeslide, iterative closest point requires you to know
              the 3D geometry of the scene to perform registration.
              altneratively, you need some kind of parametric transformation
              model for your images.
2004/1/2-5 [Consumer/Audio] UID:11649 Activity:nil
1/2     Have the musical programming merits of XM vs. Sirius been
        debated on the motd recently? If so, would someone be kind
        enough to summarize? Otherwise, your humble opinions would
        be much appreciated. Most reviews of satellite radio programming
        I'd googled had been limited to technical interest sites and the
        talk boards of XM/Sirius loyalists (who often troll each others'
        discussions) until last Friday, when one of the NY TIMES pop
        music reviewers wrote a paen to XM DJs that amounted to
        1.5 pages of the kind of positive advertising money can't buy.
        Question: does XM programming really represent a rebirth
        of the radio DJ? is Sirius not that far removed from Clear
        Channel-controlled FM radio? The input of those who play
        instruments and/or listen deliberately to music would be
        most appreciated. --elizp
2004/1/2-5 [Consumer/Audio] UID:29740 Activity:nil
        1.5 pages of the kind of positive advertising money can't buy.
        So, does XM programming really represent a rebirth
        of the radio DJ? is Sirius not that far removed from Clear
        Channel-controlled FM radio? --elizp
                \_ I think this varies from channel to channel. I can attest
                        to the much higher program quality for a few XM
                        channels I tried.
Berkeley CSUA MOTD:2004:January:02 Friday <Thursday, Saturday>