For the love of pubsub
I have fancied publish-subscribe middleware(pubsub) architectures for quite sometime. When we were still very new in CTS, I came up with a wild suggestion to try this for getting a data dump from a legacy system into our moodle instance. I was quite clueless about what this would entail in terms of development effort, but I could create a PoC setup within a few days which made me quite confident of being able to pull this off.
In fact, I had enough spare time in hand to actually implement a PHP module that would talk to an IBM MQTT server over pubsub protocol. I should have written about this experience because it was one of those times where I really perfected my mallocs and frees! The other instance was when I wrote a PAM module for authenticating against AD. Now there are many supported libraries to do this, but when I started out there wasn’t much. OpenSUSE had some customizations of the LDAP login module to support AD. Most of us in the team were pro-Fedora and hated OpenSUSE because we thought it made Linux look like Windows. In fact, I was pro-Debian and hated most rpm distributions. I still do in spite of the fact that I learned to use a Linux machine on a RedHat distribution! And I still hate Ubuntu for similar reasons. Too many digressions!
Yeah, PHP was quite pedagogic about my code. I abandoned the module because I was not clear about the licensing terms for using the C API from IBM. Perhaps the code is still lying around in one of those machines we used back in the day. And I didn’t want to bug my manager’s time to pursue the required clarifications. I am not even sure he knows about this.
So, yes, I fancied pubsub architectures ever since. If I had a chance, I might have used it in my MS thesis as well. Then I moved into Semicon jobs and haven’t really had a chance or reason to get back to this topic. My next fascination is the Arduino. About a year or two back, a bunch of people came up with mqtt implementations for Arduino. I guess it is pitted to be one of the contenders for IoT implementations. This rekindled my interest in the topic and I started reading up again. Toying around with the libraries isn’t going very far. I needed a concrete premise to play around with.
I am good at web development. I prefer to use python and am reasonably stuck with django as the framework of choice – even if all I want is a todo list. I am no django-expert, but I can swim around with reasonable ease. This also helps because I still haven’t gotten my head around any sort of nosql concepts. I briefly dabbled with the idea on GAE and pissed off a mentor! So I pretty much stick to mysql/sqlite with django which suits the biggest of what I have ever done in this platform. Nope, not a todo list, a proper report generation tool for a friend. Wait, that reminds me, I should see if I can get some inputs from him for my portfolio!
More searching lead me to redis. I don’t know enough to talk about redis, but somewhere redis seems like the perfect solution – especially because someone figured out how to use python with redis_,_ django_, and websockets and not die halfway wondering what we started out to do. Oh! I have a premise as well._
I’ll abruptly stop this post here. I’ll continue talking about this experiment and my learnings in a future post. Of course, I’ll talk about the premise as well there.