[V]andemar.org

I had my Wolverine Trading phone interview Friday, and as expected, it went well. It seems most of these initial interviews are just to make sure you’re not a psycho, and can interact with a human being without breaking into tears. Less than a minute after hanging up the phone, I had an email offering a more extended interview over the course of a day downtown. They want me to bring a code sample and my driver’s license. Both are a bit worrying, as most of the code I’ve written is modifications of existing software which I can’t claim I really wrote, and they won’t even let me buy booze with my expired drivers license, so I can’t imagine they’ll let me use it to prove my eligibility for employment. Guess I’ll find out…

Shortly after receiving my Wolverine offer, I got a response from Microsoft. They want me to come in for the “psycho filter” interview, and I’m going for it. Even more surprising, they ignored my request for a coding position, and have recommended me for a project manager position. Oddly enough, I think my skill set fills the requirements for this job better than that of a rank and file coder. I guess we’ll see if Microsoft agrees. If so, expect a damn sight better code quality in my division… If nothing else I’ll come out with a couple of new logic questions to mull over. If all goes well you’ll see me at the right hand of this man in the near future:

I’ve been continuing my work with Python and Myghty, and both have continued to impress me greatly. The last time I dabbled in web design and web services, HTML was a horrifically ugly amalgamation of content and presentation, and writing code to generate it was liking trying to herd cats. Until recently, I thought nothing had changed. After the last couple of weeks relearning HTML, learning CSS, and seeing the power of Python in a production environment, I realized how wrong I was. Though not perfect, writing HTML is actually enjoyable, and programming a web framework is even more fun. Call out the nerd patrol if you wish, but it’s the truth. I’m having a ton of fun messing around in Myghty’s internals, and designing an architecture for the Threshold project (of which I’ll explain more later, when we have an actual prototype to play around with). Python truly is an amazing language. Polymorphism runs throughout the entire design of the language, and it shows. 95% of the time, you don’t even need to think about implementing polymorphic behavior, it happens naturally as a result of the way you express classes in Python. Unlike C++, there’s no need for virtual functions, since every function is a virtual function, and can be overridden as such. A class hierarchy just seems to flow together, and fighting the language to get the structure you want is the last thing on your mind. The only real “problem” I’ve found is that Python tends to place into stark contrast any poor program architecture choices. If you mess up the design, it’s easy to see (and, thankfully, nearly as easy to fix).

The power of functions simply being an identifier and an object in Python is nearly impossible to overstate. Want to generate a list of completely dynamic functions, in the middle of execution, on the fly? Done, in a few lines of code. It’s trivially easy to develop a program which will adapt to nearly anything passed to it. Dictionaries (know as hashes in most other languages), are also exceptionally powerful, and having them as a very fast built in type is a great thing. Frustrated by the lack of a switch statement in Python? No problem. Just use a switch statement implemented through a dictionary.

result = {'option1': function1,
'option2': function2,
'option3': function3,
'option4': function4}[value]()


In another tech sidenote (this one’s a bit shorter, I promise), I submitted my first really serious bug report for an open source product. You can see it here (Suse’s bugzilla seems to be down recently, so the link might not work). Although I thought I had tracked down the memory leak in Suse 10.0 to Beagle, it turned out it was actually a kernel bug in the orinoco driver that resulted in it allocating, and not releasing, ’size-64′ objects from the slab. I compiled a custom version of the just released 2.6.14 kernel, using a nearly identical config to the kernel Suse had provide (thank you /proc/config.gz), and the problem has gone away completely. Suse’s version of mkinitrd is actually really neat. It automatically detected that I had just compiled a new kernel, what modules I needed to boot my system, packed them up, and put the appropriate initrd in my /boot directory. The kernel worked perfectly on the first boot, and I haven’t needed to restart the machine since. I’ve left a detailed breakdown of the problem on the bug report I submitted to Suse, as well as an explanation of the solution. Hopefully they include a fix when they next issue a kernel patch. Regardless of whether they do or not, an uptime measured in months, here we come!

This weekend has been a great end to a very stressful week of midterms and interviews. I finally caught up on my sleep, enjoyed the great weather, and even had some time to read a bit. I can’t help but thinking of this as my last quarter here, even if I do have to take some classes next quarter. Although I didn’t do much work for class (ECE 357 especially), it shouldn’t matter, since things next week won’t be quite as crazy as this past week. Up until Thursday that is. Here’s to hoping Motorola and I get along as well as we did during the research presentation…

~ by Andrew on October 31, 2005.