Albert White's Weblog

Downloaded the SunRay3 software for linux on Friday night with the intention of getting it working over the weekend. It didn't work straight out of the box on my SuSE 9 machine so I thought I'd upgrade to 9.1 with the online update tool in yast2.

Now I have never done this before. And it's only recently that I've started using Live Upgrade to update a solaris box, I have an automated lab to test the Solaris patches work with Live Upgrade (and patchadd -R) so my confidence in the product is fairly good. The mistake I made was thinking that an automated update would be trivial on linux.

I pointed the tool to a nearby mirror, and set it downloading. When I came back hours later I discovered it had completed! However it had mentioned it neded to download several gig, and unless my broadband connection had got ultra fast, there is no way it could have finished. So I started it again, and sure enough it went off to get more packages. I started the update 5 times in total before all dependencies were met.

I then had a dilema. An update that hadne gone too well, had updated the kernel to 2.6.something, and god knows what on the system from the previous 2.4 kernel (sorry but linux kernel modules are about as forward compatible as pouring horse feed into a petrol engine, my laptop on the other hand has a solaris 8 network driver working happily in solaris 10), should I reboot? Ah sure why not, worst that can happen is that I have to recompile the nvidia drivers right?

nvidia drivers didnt work. And didn't compile. ok my own fault for not downloading the ones for the 2.6 kernel (thought of horse feed cross my mind again). downloaded compiled and insmodded fine. reboot. panic. reboot. hang. reboot. roll a d20 to see whether it hangs or panics while setting up eth0.

Turns out that the VPN client module was making all of the networking parts of the kernel break. This module was rebuilt, with the 2.6 kernel, compiled fine, loaded fine. Then does random things, when you try to do something _else_ with networking, like run ifconfig. Found a module for the Java Desktop System and it works, but I'd have expected the broken module to complain at compile or insmod time if something was amis. It worked on 2.4 afterall.

Networking now works, with the forcedeth driver rather than the nvnet driver. According to yast2 it should be nvnet, according to modprobe.conf its nvnet, but something still insists on using the forcedeth driver. I think Linux is trying to be clever somewhere, but I'll settle for using this for the moment it works. I haven't gone near audio, scsi, scanner, printer, but based on my previous experience (not counting this weekend) of the Java Desktop System and SuSE that should be ok.

That, dear friends is the story, much abridged, of how my upgrade went at the weekend. Does the SunRay work yet? Well no (its only listed as supporting SuSE 8, not 9.1), thats another days work, but when it does I'll post pictures of the PC in the attic and Neverwinter Nights running on the sunray :-)

In other news Prof. Fred Whipple passed away aged 97. He is probably best known for coming up with the `dirty snowball` theory of comet neucleii[1], in 1950 the idea of the comets being mainly ice and having tails of gas heated by the Sun was a totally new concept. Almost as if the universe decided to mark his passing, two new comets were discovered by Amateurs last week. 2004/Q1 (Tucker) and 2004/Q2 (Machholz) may reach naked eye visibility sometime around December or January.


[1]http://adsabs.harvard.edu/cgi-bin/nph-bib_query?bibcode=1950ApJ...111..375W&db_key=AST&high=409ec4d6b916472

I own an LXD55 10"SN telescope from Meade Instruments. It's a fine telescope, big mirror, decent enough mount (just), goto, tracking... However it is on the cheaper end of their catalogue thanks to the Schmidt Newtonian design - spherical primary mirror with a corrector plate at the front, this is cheaper to manufacture than their more popular (and expensive) Schmidt Cassegarians.
However the cost cutting didn't end there. The gears backlash a bit, one is made form aluminium, the keys to hold the gears in place are not adequate and cause slop. So even though you have nearly EUR2000 worth of a scope it can be rendered useless unless you put the effort into maintaining it. I've now replaced all the dodgy or annoyingly placed bolts, and the scope performs much better. At a cost of an extra $40. I would much rather have paid Meade the $40 dollars to get these details right than to do it myself!
As a colleague once said "If I pay a million bucks for a machine, I expect to be able to swing out of the door!". With a 15k you can, I have tried, and it works, though I'm a light guy! I'm not advocating that you repeat this experiment by the way. Now I don't spend that kind of money on telescope kit, but I still expect the parts to be made to last, if that means a few dollars more onto the price tag to getter a better grade steel bolt then fine!

A friends blog mentions a NewScientist article relating to a tribe that apparently cannot understand numbers beyond three. The conclusion by the researchers is that language determines thought. Although i don't have training in this field, I don't believe the conclusion for three reasons:

1.The tribes understand three. So their thought works on "one, two, three, many" even though their language only describes "one, two, many".
2.If language determines thought, how does language evolve? These new words come from somewhere, many are combinations, many are not.
3.Thingamajig. The universal word for something there is not a word for, yet. Similarly many of us, when writing blogs for example, have trouble putting thoughts accurately into words. If language was determining my thought, this would not be a problem.

Listed in the new words above is "edutainment". I would like to add a similar word to the English language "Educution" a combination of education and execution, defined as the process by which evolution advances with a sarcastic touch of letting the soon to be deceased exactly why they are being removed form the gene pool. Some reciepients of darwin awards could be said to have been educuted.

JCDM points to the above research as an example of Bad Science, and repeats the mantra that many scientists will have been told "Correlation is not the same as causation"!

I ran into this recently in the strange field of archeoastronomy. This is a field of archaeology concerned with ancient astronomy. Was Stonehenge an observatory, are the pyramids an observatory? etc. Here we clearly see where the mantra of "Correlation is not the same as causation" needs to be applied.

For a given stone circle/megalith/passage grave I guarantee that you will find a significant lunar or solar or planetary alignment on a specific day. Given enough bits of stones (and imagination) you can come up with some extraordinary possibilities, for example that Stonehenge was an ancient computer or able to predict eclipses. I'll leave the eclipse predictions for another blog, but lets assume that it is possible to predict eclipses at Stonehenge, and that it can be used as a computer of sorts. Does that mean that it was?

Just because we have a correlation (what the people of Stonehenge could have done) does not imply that this is what they did. Lets take a techie example, MIThenge http://web.mit.edu/planning/www/mithenge.html. Imagine in future archaeologists arrive to find an intact MIT campus. On Nov 11th the Sun shines through corridors much as the way it does through the passage grave at Newgrange. Nov11th is remembrance day. Correlation... only! MIT wasn't built with commemorating the WW1 dead in mind.

"Correlation is not the same as causation"! You need more than correlation before you can state a cause.

This has implications for IT also. You may be able to correlate certain events, but not find the casue. For example say after you install a patch cluster, something stops working. Though the cause may be to do with a hardware fault, or configuration change, that did not trigger until the machine was rebooted, the obvious correlation leads you to blame the patch even though it wasn't the cause. I have seen several cases where it turned out that the root cause of the customers problem blamed initially on a patch was actually a hardware fault.

Today I went on a customer visit, a very rare thing for me. A few years ago Sun Ireland gave out some hardware to computer societies in various Universities around the country. As one of the first members of such a societiy (timf was a founding member) let me tell you a little about it.

In 1995 our university (www.ucd.ie) had no email for students. It had web access for students for about a month before realising that students were actually using it and this was crippling the college bandwidth. And there was very little unix material on the course. The UCD Internet Society was founded to help students out by educating them about the Internet, give them an email address and for those who were interested, to learn about UNIX. This was accomplished for 500 students with a 100MHz PC running a 1.3 Linux kernel, if I recall correctly the scsi controller in the machine was not supported on the more stable 1.2 kernel. People queued from early in the morning to be a member, normal students not just the computer geeks!

At its peek the society had over 3000 members, making it the second largest society in the college. That many paying members means you can buy some decent hardware, not to mention beer for the many social events!

The admins of this and other colleges were able to get real experience on unix machines. Not just the kind of experience you can get from a bedroom PC, but some understanding of how to consider users, and how to deal with the network owners (bureaucracy of the university IT department). Useful skills for sysadmins!

From these societies, Sun Ireland has taken on many interns and many now full time employees.

Early on some of the colleges decided they would like to expand their hardware and get hardware from Sun. The initial offering of computers were fairly modest, SPARC Station 5's etc., but a few years ago Sun gave out E450's to these societies. Most of these are still in use, and many of the student admins took the machine in their stride learning what they could about solaris and the machine, after all it looks good on a CV (US:resume). Some admins inexplicably decided they wanted to put linux or *BSD etc. on them, not that I have anything particularly against linux or *BSD but when Sun gives you a machine, which was qorth quite a lot in around 2000 when these were given out, you would think that running solaris would be obvious, even if just to stick on your CV!

Anyway, one of these machines fell into some disrepair, so this morning myself and two colleagues went out to rebuild the machine, give a presentation on adminning the box, and show off sun sunrays which we brought out.

Folks seemed impressed by the sunrays. They are perfect for a university environment. For example... Many universities have libraries with terminals for accessing the catalogue, these typically are powered by a noisy PC locked in a press and tucked far away from the reading areas. With a sunray there are no moving parts, so its silent. My college library had about 30 terminals (3 floors, 10 per floor) - that can be powered by a modest enough machine, since all the sunrays will be using is a web browser to access the catalogue. So you go from 30 PC's to 1 decent spec PC and a load of sunrays - hardware and replacement cost plumets (unless you want a big server!). Given the choice wh would anyone deploy 30PC's anymore. Now if you have 2000PC's on campus plus servers...

Sunray 3.0 software is beta and available for download http://wwws.sun.com/software/sunray/beta/ SOlaris SPARC and x86 and Linux, though you will nead a sunray client to really use it! Some case studies http://wwws.sun.com/sunray/success.html

Finally got around to taking my new car to work today. Its a 2000 Nissan Almera Hatchback. Described as "boringly reliable" by Top Gear, which suits me just fine. However, I should have been driving it all week, but that didnt hapen due to a clutch failure on the way to a wedding on Saturday. But now with a new clutch its fine!

So how was my first drive? Well the accelerator now does what its supposed to, ie move the car not up the revs, like it was doing while trying to warn me of imminent clutch failure on Saturday. The power steering works, found that out on the first roundabout, my previous little car didnt have power steering. And the clutch smells like its ok, can you tell im patranoid about clutches now? I'll probably set a record for how far you can drive in a suburban area without changing gear on the way home! Dublin radio was even worse than usual this morning, so I ended up listining to Cantonese language CD's, I can now say "I can't speak Cantonese", these CD's really leave your learning on a high! Ah well at least I can do more than buy mangos now.

So anything interesting in work? Well yes, I found out an interesting way to reset the admin password on a B1600. And now I'm trying to get RedHat installed automatically on it. All other machines in the lab work off some scripts we wrote around jumpstart, coming up with a similar method for linux is proving a little trickier than I had thought, making progress though, all problems seem to be related to the client not being able to mount some filesystems over NFS from the install server.

Finally a fellow amateur astronomer came across pp3. Its a very handy and fairly configurable star chart generator designed for publication quality basic star charts. It's a c++ program that generates tex and eps, which you can import to gimp to get png/jpeg files. Its not great if you want detailed accurate charts, but this is not what it was designed for. For a nice simple reasonbly detailed constallation map this is the best I have seen. For the detailed maps use xephem.