Why on earth does a book that cost $19.77 on Amazon, cost R303 in South Africa? $19.77 = R131. Even adding shipping and customs etc. should not add up to R303!!! It is a small book - not heavy, not large. How much can shipping costs be?

Amazon (USA):

Exclusive Books (South Africa):

Electronic stuff that is imported usually works out never more than the $ amount * 8. So if an Apple MacBook Pro 17" costs $2800 in the states, it costs R22000 in South Africa.

My question is - why do imported books cost double that of imported electronic equipment?

Look at the nice ad I got - and recognise the CAPTCHA style of the contents. So precious! I love it!

So I started investigating myself. The fact that the appointments are out by exactly 2 hours implied to me it has something to do with timezone issues, since in South Africa we are at GMT+0200 - also known as SAST. So I opened up the .ics file in TextWrangler and started analyzing it.

The relevant line was this one:

DTSTART;TZID="(GMT+02.00) Harare/Pretoria":20060515T120000

Reading up on the ICS Specification, I recognised the time format to mean "FORM #3: DATE WITH LOCAL TIME AND TIME ZONE REFERENCE". Which means that date and time in the example above implies 15 May 2006 at 12:00 noon SAST (South African Standard Time). However iCal imported it as 14:00 SAST.

Now I started playing around. My time zone in Mac OS X is set to "Africa/Johannesburg". In principle that should not make a difference since it is the same as "Harare/Pretoria". However I decided to test something - I changed "(GMT+02.00) Harare/Pretoria" to "Africa/Johannesburg" and reimported it. Voila - it imported just fine and with the correct time.

I have no authoritative answer on this, but my guess is that iCal does not understand the timezone "(GMT+02.00) Harare/Pretoria" since it is neither in GMT format nor in country format, therefore it defaults it to Africa/Abijan (which happens to be GMT). Thus the offset error.

Now to automate this... I need to have the least amount of inconvenience to get my .ics files into iCal and with the correct time.

I started playing with Automater and an AppleScript - however it is too much effort to copy the attachment out of Mail.app, right click on it and select the automator action. So I created a new Folder on my desktop, enabled folder actions on it and attached this script to it:

property extension_list : {"ics"}

on adding folder items to this_folder after receiving input
	repeat with i in input
		try
			set text item delimiters to ":"
			set fileNameOnly to last text item of (i as text)
			set text item delimiters to ""
		on error
			set text item delimiters to ""
		end try
		
		set the item_info to info for i
		if (the name extension of the item_info is not in the extension_list) then
			return
		end if
		
		set fixedMacFileName to (":private:tmp:" & fileNameOnly & ".fixed.ics")
		set fileName to (quoted form of POSIX path of (i as string))
		set fixedFileName to (quoted form of POSIX path of fixedMacFileName)
		do shell script "sed -e 's@(GMT+02.00) Harare/Pretoria@Africa/Johannesburg@g' " & fileName & " > " & fixedFileName
		
		tell application "iCal"
			launch
			open file fixedMacFileName
		end tell
		
		tell application "Finder"
			try
				delete file (i as string)
				delete file (fixedMacFileName as string) of startup disk
			on error the errormessage
				display dialog errormessage
			end try
		end tell
		
		tell application "iCal"
			activate
		end tell
	end repeat
	return input
end adding folder items to

Now I simply drag the .ics file from Mail.app onto the folder and it gets imported into iCal just fine.

(WARNING: Adult content following - only read rest of article if you are older than 30).

Now it is just the frustrating wait for South Africa's resellers to get stock...

* Take out battery for 10 seconds
* Reinsert and open device
* At blue NOKIA text, press and hold Ctrl+Chr+Shift+F until the Format dialog is displayed
* Select Format.

Your device will now be reset to factory conditions.

Furthermore - the screen does not tilt back far enough to my liking. It is fine if you mount your notebook on those plastic platforms - but not if you work with it on a desk like I do.

So I decided to return i for a refund. Will see how that goes. But I am NOT impressed. Will wait for the 17" MacBook Pro.

I just bought a new Apple Cinema 23" display for my brand new MacBook Pro 15.4" notebook. I eagerly unpacked them and connected them up. I almost fried my eyes with the 23" LCD panel - its brightness was just way too high. So I adjusted it to its lowest setting. Lo and behold - at the lowest setting it is as bright as my MacBook Pro at its highest setting. For any other environment than a public office building with a bright window behind you, that is just too bright to work comfortably with. And I have 4 dead pixels. So I spent R11,000 on a display that is too bright to be usable for extended periods and has 4 dead pixels. Fine - I can live with that... For now.

So I started prepping my MacBook Pro - copying 3GB of data across the network via the 1Gbps LAN from my G5. Whilst waiting for that process to complete, I started installing XCode. Then I remembered while XCode was installing and the network copy ongoing, that I needed to check something on the Internet. So I opened Firefox. Just after I clicked the icon in the doc the UI froze - I could move the mouse cursor, ping the machine, even ssh into it - but the UI was dead. Waiting for the network copy to finish did not help either. Still stuck. So after trying to kill the offending process(es) I resorted to hard rebooting the machine. Fine - I can live with that.

I started all my applications again and continued installing and customising the new notebook. I quickly went to fetch a cup of coffee to try and calm me down. Just as I was sitting down and clicking on NetNewsWire Lite to read the new RSS feeds, I received the ever-welcome MMSOD (Multilingual Mac Screen Of Death). Just like that. By clicking on NetNewsWire's icon in the Dock. Fine... I rebooted. Before the grey boot screen disappeared, the system died again with the same MMSOD. Fine. So I tried a second time to boot - now it just hung at the grey boot screen. Finally I decided to pull the AC cable, taking out the battery and performing a PMU reset. Rebooting saw a glimmer of hope approaching as the friendly welcome screen is displayed. Cool. Issue fixed. I can live with that...

So I decided to start on my Web Services project. After having countless stability issues with my AMD XP 64 3200+ system (such as CPU reaching temperatures of 90C+ even with massive copper heatsink), I decided to replace it with a nice, fast Pentium D 3.2GHz. Since the machine is quite high spec, I assumed it would run most applications with ease. So I have installed WebSphere 5.0.2, MQ, DB2 and WSED 5.1. Ignoring the fact that MQ was complaining about some undetermined error and could not start, I loaded the WSED IDE to execute one of the example Web Services. Ok so it takes about 1 minute 10 seconds to launch WSED - on a Pentium D 3.2GHz with 1 GB RAM... After it launched I located an example and loaded it. So I went through the whole process of deploying the application - just to receive a "NoClassDefFoundError: javax/xml/rpc/ServiceException" error. It is supposed to be in the jaxrpc.jar file.. So I added it to the project but it still does not work. So I decided I'll browse through the Request.jsp page to see how they have implemented it. I double clicked the file and waited a whopping 25 seconds before the page loaded. The system was *not* swapping. I double checked this. Nice.

All this and I am not evening mentioning the patheticness of IBM's software. Where in your life have you seen a Install CD come with a stupid HTML page with directions on how to locate and find the setup.exe file somewhere in the file hierarchy? Or the stupid layout of WSED (Eclipse)? Nothing is logical in that GUI. At least IntelliJ IDEA has an intuitive layout... Or the fact that WSED is 3.2GB and requires a whopping 2 GB of disk space to install? I mean come on! If you are supposed to work in an IDE every day of your life as a Software Developer but have to wait more than one minute for the tool to load, and the stupid IDE taking up 2 GB of disk space - that is downright ridiculous. IntelliJ ships as a 58MB install image and that includes JRE 1.4.

What about implementing a Web Service in Java? Using Axis or similar frameworks usually require reading about 10-15 pages of information to understand how to write a basic Web Service - then 10 more to understand how to actually deploy the Web Service in a production environment. Using .NET (and no I am not a Microsoft evangelist) is how it is supposed to be done - writing a Web Service is as simple as tagging a method and a class with two annotations (attributes) and voila! No stubs, ties, skeletons, etc.

So I decided to leave the WSED frustration for tomorrow. In the mean time I thought I'd do some work on my Acer Travelmate 8204 notebook. After logging in I saw a notification window from Automatic Updates stating that new software is ready to install. So I went ahead clicking the baloon to have the updates install. Once done, I got a nice reminder that I need to reboot for the changes to take effect. Since I usually have between 10-15 applications running at any given moment, I am usually reluctant to accept these kind of options - I usually prefer to defer them until such time that I actually *need* the update and thus have to reboot. At least Mac OS X warns you before installing that it will require a reboot. So I clicked the "Restart Later" button. After working for 5 minutes a popup suddenly appeared on my screen:

No matter what you do - it always appears after 5 minutes of clicking "Restart Later". I mean does Microsoft really think I am such a baby that I need such constant reminding? I cannot tell you how annoying that message is!

So I rebooted and the message went away. Microsoft 1 - Waldo 0. Since I needed to do some .NET development as well, I tried to get Reporting Services working but to no avail. It kept on complaining about not finding an active IIS instance. So I checked it out and discovered that even though the WWW Publishing service was running, in IIS Management Console the Web sites node was not active - it had a red cross over it. I tried to start it but it failed with the most informative message I have ever seen:

I googled for it but only got vague answers. One entry suggested that it could be network related. So I performed a telnet on port 80 and what do you know - there was an answer. Even though IIS could not start. Hmmmm... I fired up TCPMon from SysInternals and located the culprit - Skype. For some strange reason Skype believes it has to listen on port 80. Don't ask why. But that is not the reason I am pissed off. It is because the developers who wrote IIS were too lazy to translate error code 0x8ffe2740 to a simple, yet obvious message "IIS cannot start right now since another application is bound to port 80". Is that too much to ask?

So after these issues I am a bit fed up. All in one day... Just a short while ago - as I was typing this blog entry - my wife came ask me to export our daughter's pictures from Aperture 1.1. She wanted to design a book of her life so far. I clicked on the smart album containing all star rated images of her, but after about 5 pictures were loaded Aperture just crashed - no error. I launched it again but only after the fifth launch did it load all the pictures properly. Aperture 1.1 was supposed to be a bug fix release as well... Sigh.

I hate computers. I believe commercial software are fundamentally flawed since the developers behind those systems are ALWAYS pressured for time - like I had one company tell me recently, they had a choice between Quality, Features, Cost and Time. Time was fixed, cost was fixed so quality and features went out of the door. Since commercial software are by definition competative, time and money will always be constraints. It is much easier to let go of quality. Features usually are also non-negotiable.

OSS (Open Source Software) are also by definition flawed. The developers are not usually constrained by time or features or quality, however they are always constrained by cost. Most OSS projects that are non-commercial have freelance developers behind it. People who spend their evenings and weekends coding. People like Linus Torvalds etc. do not count since they are being paid by commerical companies to work on the open source software - they usually fall into the category of commerical software. The problem is that if you spend your free time on a project you cannot put in consistent effort over a long duration - something that is crucial to the success of most projects. So either the software end up being buggy or major features are missing - not yet developed. Also, most smaller OSS endevours do not have the luxury of a proper QA phase - so the code quality sucks.

Look at Linux kernel - I notice every couple of days a new "stable" release of the kernel - now at 2.6.16.11. Yes I do understand it is bug fixes that goes into the .11 release, however bug fixes also need testing. So how can a bug fix be tested and included in only a couple of days?

The gist of the matter is that software will never be reliable, and it will never operate as we expect them to for as long as we design and develop software in the way we currently do. Humans can't cope with the complexity of today's systems - that is why a new version of Windows (Vista) takes 5-6 years to be developed and software such as IBM WebSphere is so uselessly slow and buggy. How will we ever test a system complex enough to handle the intercontinental commerce of the 21st century with current techniques? We need a drastic change in technology and a paradigm shift for all developers.

Till then I'll be struggling forward through the mess of other people's code.

Some people might believe that is a Good Thing - software automatically informing you about new updates. In general that is nice, however in this case I would have felt much better if there had been no holes - or at least more realistically a hole a year. Take OpenBSD - their clame to fame is

"Only one remote hole in the default install, in more than 8 years!"

So I decided to stop using the software - and since it is not trivial to write such software in a short time span, I decided to stop providing the forums at all. Sorry if this is an inconvenience to anyone, but SH.

The problem is that my library started pushing 200GB. It grows by about 100GB each year. I already have two 400GB HDD's in my G5 (RAID1), but with the OS and other applications I was running out of disk space. Since bigger hard drives are not readily available, I had to look for some form of external storage. I'll explain in this blog what my options were and why I dediced to build my own RAID array.

Another option would have been the Apple XServe RAID. But that costs R60,000 for 750MB RAID5. Yeah it is massively scalable up to insane capacities, but it is just too expensive (and noisy).

So I decided to look into building my own RAID array. Why? Because I already had a nice desktop Pentium 4 3.2GHz machine with 2GiB RAM. Since the case can only handle 4 hard drives, and ventilation sucked I decided to replace it too. Fortunately my power supply is rated at 500W - keeping many hard drives spinning requires lots of power.

Case
My requirements for a proper case were as follow:

• It had to be sturdy - it will carry lots of weight
• Ventilation should be exceptional - but quiet
• It should support at least 8 hard drives, preferably 10
• It should not be too expensive
• It should be in a normal Desktop form factor - not a server chassis

I could only find one case matching those requirements - the CM Stacker.

IDE or SCSI?
Traditionally IDE is seen as inferior, larger capacity and cheap when compared to SCSI. With the advent of SATA based IDE, things are slowly changing. In my mind the only real benefit SCSI still has is superior mechanics in the hard disk drives themselves. The ballbearings lasts longer and the MTBF of the drives are just much better. It is however changing with the introduction of the Western Digital Caviar RE2 with a MTBF figure of 1.2 million hours. And that is exactly why I bought four WD4000YR drives.

I have an intense disgust with the onboard SATA controller when using RAID. I lost all my data 3 times now with RAID1 on various SATA chipsets built in to some modern motherboards. That is why I went shopping for a robust, hardware based SATA RAID controller. My requirements were:

• Support for SATA hardware RAID5
• Native support in the Linux kernel
• Support for up to at least 8 devices
• PCI-X since I do not have PCIe

There were two cards that got my attention - the Adaptec 2810SA card and the 3Ware 9000 series. Both seemed to be excellent cards, and it was a matter of availability that caused me to purchase the Adaptec.

Putting it all together
So to put it all together, I used a Gentoo Linux installation (nothing else is really an option - Gentoo Linux is simply the best distro there is and whether to use Linux or not is not even a question). I recompiled my kernel with the aacraid SCSI driver enabled and other usual SCSI support enabled.

I then grabbed the command line utility from Adaptec's CD - this allows me to interact with the controller. This is extremely useful to interrogate it for status, or to silence the alarm when a drive fails. It is called aacapps and is available as a rpm package. Just be careful, when running Gentoo on 2.6 you will get the error when running /usr/sbin/aaccli:

CLI > open AAC0
Executing: open "AAC0"
Command Error: <The requested controller does not exist.>

That is because the /dev/aac0 entry is missing. Just copy it from /usr/sbin/aac0. You should then see:

CLI > open AAC0
Executing: open "AAC0"

and then
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=ESSW
-----END PGP MESSAGE-----