I have many years experience in computer programming.

Motorola - 1990s

I had a job rewriting the Fortran code used for testing radios to a C++ based Windows GUI application. This was some experience for me, although I had no idea what I was doing or how to write good object orientated code. Further more, limitations in the libraries used to interface with the GPIO bus, for communication with the test equipment didn’t help.

On such problem occurred due to my practise of keeping memory buffers allocated as small as possible. This is considered good programming practise. Unfortunately, the code for reading data from external devices didn’t allow for specifying a maximum buffer size, so would keep reading data until the device had finished, hence overwriting the end of my buffer, and overwriting random memory values past the end of the buffer (such as the next stack frame).

This proved very difficult to diagnose. It meant occasionally the test devices would return an unexpected value, which would cause incorrect behaviour in the application some time after the error occurred. Eventually I wrote a function that would test the string size after every potentially dangerous operation, and have it abort the application immediately if the string size was too long for the buffer. This helped significantly.

Education - 1990s

After my work at Motorola, I concentrated my efforts at VCE years 11 and 12, and then my Robotics and Digital Systems. I also did first year Informations Systems course, which I found very useful in later years on good design for Databases.

2000s

Through this time period I concentrated on my system administration skills.

Lintory - Late 2000s

In my work at VPAC I often ended up getting all the deprecated laptops that nobody wanted any more. I often gave them new life by installing Debian Linux on to them. “What do you mean it was slow? It runs Linux fine!” One of these laptops had a tutorial getting started with Django on the desktop. I was hooked. I started writing my own Django application, Lintory. This project was designed to keep track of staff assets at VPAC, including hardware and software. Most off the shelve solutions could not keep track of different types of licenses (e.g. commercial vs non-commercial use), and Lintory was designed to solve this problem. This project is still available on GitHUB, although I haven’t touched it in years.

Eventually through my work with Lintory I realized that my code was actually Python code, and I was now a Python programmer.

For more information on my sysadmin duties at VPAC, see my system administration page.

Karaage - 2010s

Around the same time I started work on Lintory, a coworker started work on a project called Karaage. Previously every supercomputer in the cluster had its own user database that had to be manually maintained in NIS (aka. Yellow Pages). This was tedious, very manual, and prone to error. As an example, when I started at VPAC I was given the username of brian, which clashed with somebody who had the same username on one of the cluster systems.

Karaage was a Django based system that was designed to help management of cluster users. It was never designed from the ground up, it was a pet project to solve a real problem we had at the time. Gradually I become more familiar with Karaage and helped install it at VLSCI. When the lead developer left VPAC, I ended up with full responsibility for maintaining it.

This was good and bad. It was good in that I was able to start cleaning up old legacy code, and attempt to turn it into more of a maintainable professional code base. Code that has automatic CI style testing and Gerrit to manage code changes. Removed nasty hacks to make the code more easy to understand.

I also installed Karaage at sites right across Australia, including Sydney, Perth and even Darwin. It was not so good in that nobody - including my boss - had a good understanding of what I did with my time. We also had the problem that while we installed Karaage to sites across Australia, we didn’t get any maintenance contracts with these organisations. So I was unable to fix known bugs in existing installations. It also meant VPAC wasn’t earning a direct income from the work I was doing, despite Karaage being an essential component of the work we were doing.

Unfortunately, a number of these projects to set up supercomputers at external sites where badly mismanaged. When somebody realized we should be working out how much these projects were costing us, in terms on employee hours, it was found we were losing money on them. Money we could not afford to loose. Then in 2015, finances suddenly found they had made an error and didn’t have enough money to continue operations. As my work was not producing any direct income, they made me redundant.

VPAC (aka V3 Alliance) closed operations at the end of 2015. The acting CEO at the time said V3 was making a profit at the time, and he didn’t understand why the board decided to close the company down.

Karaage - post-VPAC

After I was made redundant at VPAC, I gave a talk at PyCon AU 2015 on Karaage Gardening.

Since I left VPAC I have continued to improve Karaage, as part of my casual work with University of Melbourne. It now is deployed through Docker containers, and I have removed legacy code that was no longer required.

In the future I would like to move Karaage from a monolithic design to a more secure modular micro services based design.

Karaage now has its own website.

SPUD - 2010s

SPUD is a photo database that uses Python, Django, REST, Angular, TypeScript, and I use it to manage my growing collection of photos. Unfortunately my time is limited, and I am unable to give it the attention it deserves.

In the past I have given talks of SPUD at local user group meetings, although unfortunately these were not recorded.

Robotica - 2010s

Robotica is a series of Raspberry Pis throughout the house running Python3 AysncIO based code to make annoucements and otherwise try to assist looking after Autistic kids.

Talk at LCA2018 on Robotica.

MCA - 2010s

I now have a new personal project, MCA (misleading connection advisory), which helps make connections between services. I have encountered two problems with existing planners:

  • They don’t use real time information. So they are advise you to make connections that you can’t possibly meet.
  • They tend to be inflexible if transit problems occur. For example, if there are no trains past Ringwood on the Belgrave line, there are several alternative options available, but it isn’t possible to tell the planners not to use trains between Ringwood and Belgrave when coming up with a solution.

As an extra bonus, this project is written in Elixir, and uses Phoenix, so I am now learning to program in a functional programming language.

Personal projects - 2010s

Please see a full list of my current projects.