First Cut of Cap’n Hook
This morning I've finally put the docs and a few tests on my Clojure library called Cap'n Hook - after the cereal, and the fact that it's all about making web hooks easier to implement in services. When I was looking for my current position, I was thinking what could I write that would be fun and useful at the same time. One of the things that I knew were really useful were the web hooks that allowed one-way, asynchronous, message passing from a service to a set of registered clients.
While it's not really a good replacement for a real messaging system like any of the messaging offerings from AWS, or Tibco, or even Java MQ - it is very easy to implement, and if you're in the world of RESTful services, it's a really handy thing to have, as it's super simple to implement.
Also, if you put in retries, as I have, then it's pretty close to reliable - but only if your receiver is online and taking POST messages on that URL. Sure... it's not perfect, but for dealing with simple messaging between data centers, it's hard to beat. There are a ton of services (GitHub, etc.) that are using this to fire-off events on the client-side of things.
So it's pretty simple to use - the key is that for most cases, you are going to need to have some persistent set of registered URLs, These are the "endpoints" for the HTTP POST calls that this library will be sending in response to the main application saying "Here you go!".
I was a little blocked for several months on this - because I wanted to be able to implement a shared, persistent, reliable storage for the registrations - and then yesterday I realized that was just a bad idea. Each application that uses this will want to implement their own registration their way - and while I implemented a simplified registration that wasn't shared or persistent, the ideas are there, and it's easy to slot in a different registration scheme at any time.
I can see using redis to store the URLs - it has all the primitives to do what's needed. And there's Postgres - or any database, really, and I even cover that in one of the examples in the README.md on the GitHub repo. So there's no need to really solve the registration problem for everyone. Just give the users a simple one to get going, and then they can get as fancy as they need to.
I spent a little time on writing some unit tests, but it's going to be hard to really test the sending as it really needs to hit something, but I'll spend a little time today seeing if there isn't an "echo" test site out there for just this kind of testing, and then I can test sending the POST messages to see if they work.
But it's been a lot of fun to get this done and into testing... The documentation was great fun, and it just was a great little thing to finish.