Work: Getting there.. probably.

Well, a bit of a breakthrough on the undergrad lab front.

I’ve managed to get all the old and news users set up on the new server for both Windows and Linux use. Well, I think I have. The test ones I’ve been playing with seem to work anyway.

Still busy.

Work’s still really busy.

I am actually making headway with migrating the users from the Sun server to the Dell/Microsoft one. It’s a pitty that Integrix/SFU’s tar barfs on files generated by Solaris tar or GNU tar and the build environment is too broken to allow GNU tar to compile. It means that I have to use cygwin to extract the user data but cygwin REALLY screws up the NT file permissions in interesting ways. I seem to have got NIS/NFS working fine with the prototype client machine when it’s running Linux.

Somehow I’ve got to find a way of making the global group policy only apply to a certain set of machines rather than all the machines in the active directory. I have until the 11th to get these software things done as well as all the hardware to configure and work out how I’m going to image the client machines (I’m thinking knoppix + dd + NFS at the moment).

Oh, on top of this I have postgrad user accounts to set up, then do their induction on the 1st. Oh, and a new professor who has needed his new machines built, two workstations and a laptop plus two printers.. none of which have been plain sailing.. Oh, and the entry system where people keep forgetting that they need to register their cards, so I’m having to stop sporadically to do those as well.

Anyway, I’m knackered. Another early night tonight.

This week.

It’s been a VERY busy week at work. (I dunno why, but I almost typed school there.)

Basically, May, my colleague, and myself have to migrate our undergraduate teaching computer system from a Solaris UNIX server to a Dell Windows one. This includes moving all the data and the accounts.. Oh and making the Dell an NFS/NIS server so that the dual booting PCs can use the home directories under Linux. The whole thing is complicated by the fact that the Dell server will straddle two sets of IP addresses, the internal, private 10 network and our public 163.1.22 network. Add to this that the server cannot be a proper DNS primary server as the Oxford University DNS is a single, flat file with all the hosts being named host.unit rather than delegating authority and it makes setting up Active Directory nearly impossible.

After four reinstalls of Windows Server 2003 (resetting A.D. screws up the machine totally, so a re-install is necessary, it seems) I think we have a mostly working service on that front. Now “all” we have to do is get the user accounts set up properly and the data moved. This, of course, is going to be a major problem in itself. I’ve imported the password file using the “Services for Unix” migration tool, but I will need to go through each entry doing all the final settings.. Only a couple of hundred there. I just hope that it actually works.. Oh, and then I have to create all the profile directories and share them and set their permissions.

We also have to unpack and install all the 34 PCs, create the disk image on the prototype machine, including getting all the applications installed before imaging it and finally put the image on the machines.

All this has to happen in 3 weeks, before the beginning of term, ready for classes to use it for practicals almost on day one.

If you add to this the fact that we also have to service all the normal day-to-day problems in the department, induct the new postgrads who arrive in a couple of weeks, install a few new Sun W2100z workstations, two of which need to run Windows but Sun doesn’t supply the drivers and I needed to find the right ones to allow the Windows install to actually see the disks. Oh, and then there’s the new access control system with the swipe card management with software which makes it as difficult as possible to add cards.

I think you get the idea about how busy it is at the moment.

Anyway, today’s been pretty quiet. All I’ve done is shop in Oxford for wet weather gear. I’ve found a decent waterproof coat with integral fleece (Sprayway) and a day-glow yellow lightwieght waterproof jacket for cycling. The last one replaces my old cycling waterproof jacket which is a mere 20 years old exactly.. It was bought just before going to Uni for the first time. Unfortunately, its waterproof properties have failed even though it looks OK. Even after several dousings of waterproofing stuff it still lets water. I wonder if the new jacket will last that long.

Chicken or the egg?

Having read the article from New Scientist I’m left with the question which isn’t answered… Isn’t there a possibility that the people who write diaries regularly probably going to be the character type who have the medical and phsychological problems described as the consiquences of diary writing in the article?

From what I can see, the only way to prove or disprove this would be to find a large sample of non-diary writers, do a medical on them all and then get half to start writing diaries without telling them what the study is for (so as not to cause a placebo effect). Just testing those who are already writing or not writing is pointless as you can’t separate cause and effect.

Sometimes I dispair about the psychologists (and many medical researchers) who seem to be pseudo-scientists who abuse statistics to prove the conjecture they formulated before the study. A true scientist should try as hard as they can to knock down their own conjecture or theory and only if it withstands every attack they can think of start to believe it might hold some sense of truth.

64 bit madness

Well, last night I finally tried putting a 64 bit operating system on my laptop as an experiment.

I shrank the Windows partition by a further 5GB, put a new partition on it and installed the MandrakeLinux 10.0 RC1 AMD64 distribution on it.

So far, other than the 2.6 kernel which is used during the install process not being able to access the keyboard I’ve not had any problems. Of course, it can’t use the wireless ethernet as ndiswrapper only works as a 32bit driver. I built a newer Linus kernel (2.6.7) and installed that as x86-64 support is getting (generally) better with each release. I didn’t put 2.6.8{,.1} on there as that seems to be a rather buggy release.

The machine does feel substantially quicker than when it was running in 32bit mode. I think the kernel build was almost double the speed, though this is only a feeling as I didn’t time it. However, after the kernel build using the original kernel I did notice that the system started bogging down and becoming a little unresponsive but I’ve not seen this (yet) with the newer kernel.

Anyway, it looks promising.