So, as a pet project, I put together MacWhines.com. It’s a very simple voting website where you get to bitch and whine about all things Apple. While creating the website, I took the opportunity to learn a little about OpenID and decided to rely completely on OpenID for the sign in mechanism. Additionally, I added “Sign in with a Google Account” and “Sign in with Yahoo! ID” buttons that redirect you to Google or Yahoo! and allow you to sign in via their federated login systems. Give it a spin and let me know what you think.

PS. There’s not a lot of functionality on the website apart from signing in, voting for whines, submitting your own whines and commenting on existing whines, but that’s kind of the idea. I didn’t want to overshoot the target so to speak.

Fingers crossed!

The problem I encountered was that the installer would hang at the very end of the installation process. In some cases, the box would even blue screen at this stage. I scratched my head on this problem for some time, but today, I finally figured out how to solve it.

I had a look at the Windows Task Manager when the installation had hung and discovered that there were a couple of processes called spsslinit.exe that appeared to be doing nothing. So I fired up the installer once more with MSI logging enabled and when it had hung, I looked at the end of the log. There, I found mention of a script function called PostInstallCommands. From the log, I could tell that the script function started up 7 spsslinit.exe processes (with various different parameters). I don’t know for sure what the spsslinit.exe processes do, but I believe they are somehow related to SPSS licensing when running in standalone mode. We use a network license, so we don’t need the standalone stuff.

The next step was to somehow prevent the SPSS installer from running the PostInstallCommands script. So I fired up Orca (Microsoft’s MSI editor) and looked in the InstallExecuteSequence table sorted by sequence number. Sure enough, at the end of the sequence, there was PostInstallCommands. I then created a new transform and dropped the PostInstallCommands row together with another row called SPSSlinit. I’m not sure the latter is needed, but I thought it can’t hurt.

With my brand new MST file, I started the sequencer and fired up the installer with the following command line:

msiexec /i “SPSS Statistics 17.0.msi” TRANSFORMS=1033.mst;SPSS17_Softgrid_Fix.mst

It worked beautifully! After the install, I upgraded to 17.0.1 and even that went without a hitch. I’ve done some testing and the app is running just fine with our network license. If you’re trying to do this with a standalone or site license, I’m not sure this is the way to go, but then again, I don’t think SPSS can be sequenced using anything other than a network license.

I’ve uploaded the MST file if anyone needs it.

The MAC address entered is not in the valid range. Valid values are between 00:50:56:00:00:00 and 00:50:56:3f:ff:ff.

The reasons why VMware force manual MAC addresses into this scope are unclear to me (probably to prevent collisions), but regardless, it can be an issue. I needed to specify a particular MAC address in the 00:04 range. The reason being we were retiring an old physical server running the FLEXlm licensing service and the licenses were granted for the MAC address of the NIC on that machine. Renewing the licenses for the new virtual machine turned out to be ridiculously expensive so our only real option was to “spoof” the MAC address.

I then ran into the aforementioned issue. However, the problem was easily solved by setting the MAC address from within the guest operating system instead. In this case, the server was running Windows Server 2003 and changing the MAC address was as simple as modifying the parameter “networkaddress” in the properties for the NIC in device manager. Voila!

The morning after we discovered, as we logged in to a few of the servers, that all virtual machines had shut down unexpectedly the night before. My first instinct was that the network engineer must have disconnected the power to the VMware servers for some odd reason and that this was the cause for the disturbance. But after some digging around, we found that none of the ESX servers had rebooted. So why had all the virtual machines shut down?

Turns out, there’s a feature in ESX called isolation mode. When you’ve setup your ESX cluster for HA (High Availability) and an ESX server loses contact with the other ESX servers and with its gateway, the server considers itself to be isolated and the default action when isolated is to shut down all virtual machines! In other words, if you have to disconnect your ESX servers from the network, look into isolation mode first.

I’m still not sure why the default is to shut down the machines. I mean, won’t forcibly shutting down a virtual machine potentially cause more problems than letting it run without a network connection?

How does this apply to web design? Well, some clever web designer figured out that you can composite a number of smaller images into one large image and then use CSS to display a particular part of that image. The advantage is that the browser only needs to request one image from the web server even though there appears to be many on the page. This leads to fewer HTTP requests and all-round improved performance. Additionally, it decreases any kind of lag that might be introduced when switching images.

Using CSS sprites is easy. Just combine a number of images into one and use the offset parameters of the CSS background property to define which part of the image you want to display. I use CSS sprites for the icons in the main navigation of this blog. The composite image is called nav_sprites.png. As you can see, the first sprite in the image is the blog icon (the plus sign). In CSS, I’ve defined this as:

background: url(img/nav_sprites.png) 0 0 no-repeat;

The hover style for that link is defined as:

background: url(img/nav_sprites.png) 0 -28px no-repeat;

In other words, the same image, but the image for the hover style is offset by -28 pixels from the top of nav.png. The next image is offset -56 pixels and so and and so forth. It should be worth mentioning that in addition to setting the offset, you might also need to set a width or height for the element in question to prevent stuff from the adjacent sprites to be displayed.

In my case, I placed all sprites in one “column”, but there’s nothing stopping you from adding multiple columns to the master image. In that case, you need to use the horizontal offset parameter as well (0 in my case).

Personally, I love these kinds of tweaks. CSS sprites are really easy to implement and even though the performance gains might not be huge, I’m a big fan of anything that streamlines the operation of computers and technology in general. There’s been some discussion on where to draw the line when it comes to CSS sprites. One could, in theory, include all images that are used on a website in one big composite image, but I think that’s taking it too far. If the browser needs to download a huge image before it can display any graphics at all on the page, the user experience will suffer. I think a good practice is to group images together. In other words, one composite image for icons, one for general UI elements, one for borders/corners etc. And don’t forget, this only applies to images that you display with the CSS background property, not the <img /> tag.

Happy spriting!

Edit: I forgot something. If you’re a bit lazy and you already have a large number of images that you want to combine into one master image, there are a couple of generators online that will do the job for you. I’ve tried CSS Sprites Generator and it worked great. There’s also this one which appears to have more features, but I have yet to test it myself.

I’ve been playing around with the iPhoto ‘09 face detection and face recognition stuff today. Having tried a lot of stuff to get the best results, I now want to “reset” the face recognition database or whatever so that I can start from scratch and teach it properly. Does anyone have any idea how to do this apart from emptying my library and re-importing everything again? There must be a way, right?