January 2014 status
2014 07 Jan
I will now be regularly writing posts about project status, plans and hopes for the future.
Before that though, there's one important news to share.
Until a year ago all development was financed through consulting and development services. This worked alright but too much time was spent doing things that didn't benefit the open source projects. And that didn't make me happy at all. Because I like being happy I stopped that for the most part and spent the year figuring things out, experimenting and discussing with people about it.
What makes me happy is answering these "what if" questions. Ranch and Cowboy are a direct product of that, as they originate from the "what if we could have a server running different protocols on different ports but all part of the same application?"; Erlang.mk is a bit different: "this works great for me, what if it could become the standard solution for building Erlang applications?".
When I successfully answer the question, this becomes a project that may end up largely benefiting the Erlang community. I love Erlang and I love enabling people to build awesome products based on my projects. It's a lot more rewarding than activities like consulting where you only help one company at a time. And it's also a much better use of my time as this has a bigger impact on the community.
The hard part is to figure out how to be able to spend 100% of the time on projects that you basically give away for free, and still be able to afford living.
The immediate solution was getting work sponsored by the LeoFS project. LeoFS is a great distributed file storage that I can only recommend to anyone who needs to store files or large pieces of data. The sponsorship works pretty great, and spurred development of the SPDY code in Cowboy amongst other things, plus a couple upcoming projects done more recently and getting a final touch before release.
It turns out sponsoring works great. So I'm thinking of expanding on it and hopefully get enough sponsoring for fulltime open source development. So I figured out a few things that can give incentive to companies willing to sponsor.
Sponsors can request that a particular version of Cowboy be maintained indefinitely (as long as they're sponsoring). This means fixes will be backported. This doesn't include features although I can take requests depending on feasability.
Sponsors can have a direct, private line of communication, useful when they need help debugging or optimizing their product.
Sponsors can get their name associated with one of the project and get a good standing in the community thanks to this. They would be featured in the README of the project which is viewed by hundreds of developers daily.
Sponsors can be listed on this website. I will modify the front page when we get a few more sponsors, they will be featured below the carousel of projects.
Please contact us if you are interested in sponsoring, and say how much you are willing to sponsor. The goal here is only to have enough money to make a living and attend a few conferences. There's an upper limit in the amount needed per year, so the more sponsors there are the cheaper it becomes to everyone.
The upper limit stems from the new legal entity that will replace the current Nine Nines. This is mostly to lower the legal costs and simplify the administrative stuff and allow me to dedicate all my time on what's important. From your point of view it's business as usual.
Now on to project statuses and future works.
Cowboy is getting ready for a 1.0 release. Once multipart support is in, all that's left is finishing the guide, improving tests and finishing moving code to the cowlib project. I hope everything will be ready around the time R17B is released.
I already dream of some API breaking changes after 1.0, which would essentially become 2.0 when they're done. An extensive survey will be setup after the 1.0 release to get more information on what people like and don't like about the API.
And of course, when clients start implementing HTTP/2.0 then we will too.
Ranch is also getting close to 1.0. I am currently writing a test suite for upgrades. After that I also would like to write a chaos_monkey test suite and add a getting started chapter to the guide.
Ranch is pretty solid otherwise, it's hard to foresee new features at this point.
I didn't expect this project to become popular. Glad it did though.
Windows support is planned, but will require GNU Make 4. Thankfully, it's available at least through cygwin. Make, Git and Erlang will be the only required dependencies because the rest of the external calls will be converted to using Guile, a Scheme included since GNU Make 4. So it is Guile that will download the needed files, magically fill the list of modules in the .app file and so on, allowing us to provide a truly cross-platform solution without losing on the performance we benefit from using Make.
Also note that it is possible to check whether Guile is available so we will be able to fallback to the current code for older systems.
I am also thinking about adding an extra column to the package index, indicating the preferred tag or commit number to be used. This would allow us to skip the individual
dep lines entirely if the information in the package index is good enough. And committing that file to your project would be the only thing needed to lock the dependencies. Of course if a
dep line is specified this would instead override the file.
This is the two-parts project requested by the LeoFS team. This is essentially a "distributed bigwig". I am hoping to have a prototype up in a few days.
Alien is the part that allows writing and enabling probes in your nodes. Probes send events which may get filtered before being forwarded to their destination. The events may be sent to a local process, a remote process, over UDP, TCP or SSL. Events may also be received by a process called a relay, which may be used to group or aggregate data before it is being sent over the network, reducing the footprint overall.
Shaman is the UI for it. It will ultimately be able to display any event as long as it's configured to do so. Events may be logs, numeric values displayed on graphs updated in real time, lists of items like processes and so on.
That's it for today! There will be another status update once Shaman is out. But for now I have to focus on it.
As always, please send feedback on the projects, this post, the sponsoring idea, anything really! Thanks.