λ. Aaron Weiss
A Brief History of Pdgn

This is a cross-post from the PdgnCo Community Blog. It was published on February 11th, 2015, and can also be found here.

The idea for Pdgn first came in June of 2014. For reasons that I can’t recall, I was drawn to searching for a domain. I didn’t really have a purpose for it. I was just poking around to see if anything cool was available. I stumbled upon the domain pdgn.co, and I thought it was concise, oddly charming, and easily pronouncable (as the English word, pigeon). As I said, I didn’t have a purpose, and so, I didn’t purchase the domain. I did, however, keep it in my mind.

Around this time, my productivity levels had plummetted immensely. I had begun to loathe working in Java, and my constant attempts to take on incredibly large projects were going nowhere. I’d been trying to learn Haskell on-and-off for roughly a year, and I’d fallen completely in love with functional programming as a paradigm. In another IRC network that I call home (FyreChat), I’d also been exposed to Rust. I didn’t really think of myself as capable of programming in a systems language, but I liked that Rust had many of the nice idioms that I had come to appreciate from my struggles with Haskell. So, I wanted to jump ship from Java, and I was looking at both Rust and Haskell. However, I didn’t have any ideas of reasonable things to work on once I did. That changed in July.

In July of that year, as the next step in a big push to distance myself from Google, I had decided that I wanted to run a small, privacy-first email service and that I wanted to write all the software myself. I had some experience with IRC as a protocol, and I figured that the email protocols couldn’t be that much worse. So, I bought the domain. I tried to make the decision of whether I wanted to write it in Rust or Haskell. Rust would be hard because I was scared of the idea of having to manage memory myself, and Haskell would be hard because I was still struggling to understand how to work with state and the real world. Ultimately, at the urging of some friends, I decided that I would write the service in Rust. I also decided that I may as well combine it with a privacy-first chat service as well. My goal was to incorporate the best privacy practices available for existing protocols, and thus I wasn’t going to invent a new chat client or a new email protocol and so on.

Still, even with an idea, my motivation was pretty low. I looked at the task of learning a new language as an impossibly high barrier, and much like previous projects, I worried that it was too large of a task for me to finish. Having already assured some of my friends that it would happen, I continually put off the idea and then put it off again. By the time summer ended, I had made no progress at all on my goal, and had made no effort to learn Rust. My friend Jacob bugged me countless times about writing the service because he wanted a new email address himself, but even that had done nothing to drive progress.

I was about to start University, and I stopped to look back at what I had done on the summer. When I did, I was saddened to see that I had done just about nothing and I wondered why. I wrote a blog post about it, and decided that I needed to do things differently. So when I started school, I decided that I was going to learn Rust by working with something familiar before doing anything unfamiliar. Dungeons and Dragons, 5th Edition was released around this time, and I wanted to play it with people over IRC. So, I thought that it would be a good opportunity to write an IRC bot to run the game. There was a clear path to starting off small, and a clear path for it to be more complicated. So, it seemed like a great first project. I split the project into two parts, the IRC library and the bot itself, and I set off to learn Rust.

From September 10th on, I was throwing all the free time that I could muster into this bot. Bored in my data structures class, I started using that time to work on it, too. Once I got over the hump of struggling with the language (and especially lifetimes), I started making good progress. I knocked out a lot of the features I had planned, and by the end of October, I found myself looking mostly at some of the harder stuff. I wanted to implement a battle map, and that required an associated web server component. I was worried about how hard it was going to be, and so I went looking elsewhere for places to continue my learning of Rust. Eventually, it occurred to me that a part of my goal had been to run an IRC server in Rust. I obviously couldn’t write it immediately, but I could definitely launch a server with an existing IRCd and make it a long-term project.

With that, on October 27th, 2014, Pdgn as an IRC network was born. I reached out to my high school friends Jacob and Alok, and asked them to join. We had run an online computer science competition earlier in the year (HSCTF), and I had missed being able to interact with them over IRC. In what can only be described as perfect timing, PicoCTF had also started that day. This meant that Jacob and Alok, both participating in it, were immediately in contact with many of the participants of HSCTF who spent their time in our IRC channel on Mibbit during and after the competition. The channel had all but completely disappated by this time, and so I hadn’t really heard from any of them. They both took this as an opportunity to recruit, and they convinced a number of old friends (and former HSCTF participants) to join the network. Slowly, but surely, we garnered a small userbase.

Seeing all the progress that was made in a day, I immediately started work on our own set of IRC services written in Rust. I didn’t have a lot of knowledge about how they were implemented, and so, I assumed that they were just normal bots. For anyone looking to not replicate my mistake, services are almost always implemented as a separate server linked to the main hub. Regardless, I carried on blindly. Within two days, nickname and channel registration was implemented. The services were starting to shape up, and I was excited to be putting them to immediate use. One issue I encountered along the way was that the user mode marking that you’re identified (+R) is actually only able to be set by a server. I didn’t have a server component to my IRC library, and I knew that that would be a huge investment. So, I modified the m_samode module for InspIRCd to allow operators to set the mode +R with the SAMODE command. I was the only server operator, and so I figured that it wouldn’t be much of an issue. With that, the bot was able to mark people as being identified.

A few days later, I found myself joining a discussion on the Mozilla IRC about IRC libraries in Rust. As far as I knew, my library was the only one that built on the latest Rust, as many of the previous developers had abandoned their work. While my library worked fine for my purposes, others were critical of my use of callbacks to define IRC functionality. Another developer who had previously worked on an IRC library pointed me in the direction of a better design. They recommended that I take advantage of iterators because of all of the sugar associated with them in Rust. So, on November 2nd, I did a large refactor of my IRC library. I dropped a lot of the excess, and implemented an iterator-based design. From there, I started down a long path of improving the library. I wrote a collection of utility functions that evolved into a utility wrapper to the server objects. I rewrote tons of unit tests. I added SSL support, and working user tracking with access level support. I dealt with crate name squatting on the Rust crate repository, and eventually claimed the crate name irc. I made the library thread-safe, and fully specification compliant. The library grew into something substantial, and I was happy for it.

Both bots weathered the storm of the redesign, and while the Dungeons and Dragons bot had stagnated, the services bot continued to grow and expand. At Jacob’s urging, I implemented the game Resistance as an optional feature for it. I also added a counter to track stupid mistakes, and a full-featured voting-based administration tool. The idea was to use the bot (named Pidgey, and declared our mascot) to allow fully democratic channel administration. We found out quickly that this was less than desirable. People started lots of non-sense votes, and rarely did votes ever pass. Eventually, I retired the democracy feature, and Pidgey went back to just managing channel and nickname registration (with Resistance and derps on the side). The server kept on running.

After a few months, it became more apparent that running an IRC network on a single server was less than desirable. I wasn’t able to do updates of any kind, and maintenance meant that everything was completely inaccessible. So, I set out to make Pdgn into an actual network instead. The first step was to move my site off of Ghost, which was being hosted on the same server as the IRC network. Once that was done, I got two new, smaller servers for the network. One in San Fransisco, and one in New York. I had to decide on names, and I wanted an overarching theme for them. So, I settled on genera of pigeons as an appropriate name. The hub server in New York was named Columba, after the genus of typical Old World pigeons. The server in San Fransisco was named Raphus, after the genus of the dodo (which is, to some people’s suprise including my own, a type of pigeon!).

On February 3rd, 2015, both of the new servers went live, and the original server that housed Pdgn was taken down. With this, the original services bot was also retired. On February 9th, I released an official pdgn.co site, and then on February 10th, I released the official pdgn.co community blog. This brings us to today, February 11th, where I have now, for the first time, documented the history of the network. It’s hard to say where the future will take us, but I hope to expand the network with more servers and more people. This is really only the beginning.

Comments? Questions? Email or tweet at me.