Nine Nines

Cowboy 2.0 release candidate 2

Wed, 23 Aug 2017 18:00:00 +0100

Cowboy 2.0.0-rc.2 has been released!

This is the new recommended version of Cowboy. Its API should not change before release. While you probably should not use it in production yet, many do successfully. Use at your own risk.

This new version contains fixes for the following issues:

HTTP/2 server push was using the wrong header compression context.
HTTP/2 flow control could end up queueing data in the wrong order when resuming the sending of data.
The shutdown timeout was not implemented and request processes could be killed too quickly.

What else is new in Cowboy 2.0? Check the rc.1 announcement.

Please test this new release candidate and provide feedback! The best place to provide feedback is via tickets. Do not hesitate to open a ticket or bring an existing ticket to my attention to get it resolved before 2.0.

Cowboy 2.0 release candidate 1

Mon, 24 Jul 2017 18:00:00 +0100

Cowboy 2.0.0-rc.1 has been released!

This is the new recommended version of Cowboy. Its API should not change before release. While you probably should not use it in production yet, many do successfully. Use at your own risk.

The plan is to have a new RC version every couple weeks until the summer ends or later if there are still blocking issues open. Only issues that can’t be fixed without making breaking changes to the interface may block the release.

Numerous changes were made since Cowboy 1.0. The one most people care about is probably HTTP/2, but it’s most likely not the only one worthwhile. The full list of changes (excluding types) can be found in the user guide chapter Migrating from Cowboy 1.0 to 2.0.

As this is not the full release just yet, please pay extra attention to details and report anything suspicious you find.

There are still some tests failing. Most of those are related to standards that are not being followed perfectly just yet. Some of those will probably not be fixed before Cowboy 2.0 is released. They are edge cases and should not prevent interoperability.

The manual received a lot of love. It now has one page per function with a detailed description, arguments list, return value, changelog and examples. It also links to the other relevant manual pages: https://ninenines.eu/docs/en/cowboy/2.0/manual/

Cowboy is now tested and supported with Erlang/OTP 19.0 or above on Arch Linux, FreeBSD, OSX, Ubuntu and Windows 7. Contact me if you can provide permanent access to another platform for the purposes of testing.

Cowboy is now available from four locations:

They are updated at the same time so there is no real difference.

The elephant in the room

Sun, 26 Mar 2017 00:00:00 +0100

Have you ever tried telling someone why they should use Erlang? You boast the smaller code size, the auto healing mechanisms, the distribution and they seem really excited. They wonder why they never heard about Erlang before. And then you show them what the code looks like. All excitement goes away. The smiles disappear. Their face starts becoming really serious.

You lost them. You know you lost them. They comment on the syntax, or perhaps you do, already admitting defeat. It’s unlike anything they have ever used before. And they will most likely end up not using it.

What about people who already know what the syntax looks like? As soon as you mention Erlang, the topic of the syntax comes in. It’s like nothing else matters.

Perhaps the topic of syntax didn’t come up. But they’re still not going to try Erlang because of it.

You’re probably not having these kinds of interactions at Erlang conferences. This doesn’t happen with people who are already somewhat interested in, or need, the features that Erlang provides. With them the syntax is at worst a minor inconvenience.

This happens because most developers are familiar with syntaxes that look nothing like Erlang. To be clear, I include language features and other concepts like objects as part of "syntax" here. Familiarity is a very important factor to drive adoption.

You can see an example of that in the Elixir world, where the majority of people come from Ruby or already knew and liked Ruby. The 2016 survey tells us that 59% of Elixir developers were using Ruby primarily before. That’s in large part because of the syntax. They will deny it of course and find other reasons. And yet, we don’t see such a strong adoption of Erlang from Ruby developers, before or after Elixir appeared.

Side note: have you ever wondered why the Elixir community is, I quote, much friendlier than the Ruby community? Despite having much of the same people?

Before we continue, let me be clear. I love the Erlang syntax. It is simple and explicit. It is powerful, especially when dealing with binary data. It has very few quirks. It has little to no ambiguity. It’s great. Except for persuading people to use it.

Over the years I have been writing Erlang, I have seen very few people point out that the syntax slows down adoption. We have no problem with it, so why would others? At the same time, people coming to Erlang come to solve a real problem they’re having, so the syntax is fairly secondary. Even if they hate it at first, they know they can solve their problems despite the syntax.

You don’t build a popular product or language by solving people’s problems though. In general you end up solving some problems and creating new problems. No, you build a popular product by convincing people to use it. And you make them stay with your product by making them commit to using it.

Take MongoDB for example. It didn’t become popular by working, or even by being practical. It wasn’t performing its primary function and was losing people’s data. That didn’t stop it from becoming popular. Smart people would knowingly use a database that was losing data. Think about that for a minute.

MongoDB of course had a huge marketing machine, and they focused on that. They helped organize many meetups all over the world, complete with various swag items given for free, including a small handbook about MongoDB. All people had to do was show up.

They didn’t go tell people to look at all the weaknesses their product had. They focused on the strengths. On what would convince people to try it. They would go to meetups, discuss with others, commit to try it (or try it at meetups directly), and by doing so sell MongoDB to themselves.

How do we get people to meetups though? That’d be the first step: you need to catch their attention. I believe MongoDB did this using benchmark results. Ironic isn’t it? MongoDB gets fast benchmark results because they lose data, and this gets everyone to buy into the product.

The key points to remember about this are:

catch people’s attention
show your product’s strengths
make people take a commitment

Once they commit to something, you win. Everyone will not end up ultimately using your product of course, but it’s at the very least become a consideration. It’s on their mind. Their resolve will be stronger when they ultimately try it and inevitably run into issues.

Erlang’s syntax is a weakness. Almost nobody looks at the Erlang syntax and falls in love with it at first sight. No, it takes time to learn it and understand how good it is. You need to sell Erlang to people without showing the Erlang syntax. If you do show it, then you need to hide the parts that feel alien. Function calls are OK. Recursion, not so much. Maps are OK. Records, not.

Avoiding code is not always possible when you try to sell it, especially to developers. You can however prepare them to accept the alien syntax by admitting that the syntax is not perfect before you show it. You can do this while praising it at the same time. For example, "the syntax is a little out there, but it matches the concepts perfectly, it will all make sense when you start learning".

This might not be the best introduction. Someone will need to A/B test it to find the one that gives the best results. But that should give you ideas.

When something terrible happens, mentioning that this isn’t the end of the world before you tell others what happened will soften their reaction. When someone breaks your favorite item and cries over it calling themselves stupid, it’s harder to get mad at them, compared to the same event with no emotional reaction.

Our behavior is largely dependent on what’s at the top of our mind, so it’s up to you to take advantage of this to make your case in the best conditions.

Next time you try to make someone use Erlang, remember that you should aim for getting a spoken commitment out of them, if possible before you show the syntax. If that’s not possible, then prepare them to accept the flaws or the weirdness before they see them.

Don't let it crash

Sun, 22 Jan 2017 00:00:00 +0100

We have a specific mindset when writing Erlang programs. We focus on the normal execution of the program and don’t handle most of the errors that may occur. We sometimes call this normal execution the happy path.

The general pattern behind writing only for the happy path, letting the VM catch errors (writing them to a log for future consumption) and then having a supervisor restart the processes that failed from a clean state, has a name. We call it let it crash; and it drives many of our design decisions.

It’s a really great way to program and the results are fantastic compared to most other programming languages. And yet, let it crash barely convinced anyone that they should use Erlang. Why would that be?

You may already know that Cowboy is capable of handling at least 2 million Websocket connections on a single server. This is in large part thanks to the capabilities of the VM. Still, 2 million is good, much better than most other servers can do.

Cowboy is not just a Websocket server; it’s also an HTTP and HTTP/2 server, and it handles many related features like long polling or the parsing of most request headers.

Can you guess how large the Cowboy codebase is, without looking at the source?

Do make sure you have a clear answer in your mind before you go check.

Good, you are back. Now what were the results? If I am correct, you overestimated the size of Cowboy. Cowboy is in fact about five thousand lines of code. You probably thought it was at least ten thousand. About eighty percent of readers will have overestimated the size of Cowboy. And you did only because I mentioned it can handle millions of Websocket connections.

Numerous studies show this effect. Just mentioning the large number already prepared your mind to think in that direction. Repeating the number made you focus even more on it. Then the question asked for a number, which ended up larger than the reality.

The same effect can be applied to negotiation for example. You generally want to start by giving your offer (and not let the other party initiate) and you want to give a really large number first. You can also prepare your customer by mentioning an even larger number in the previous discussion.

And it’s not just numbers either. An experiment showed that just by looking at an image of clouds, customers of a pillow store were buying pillows more comfortable (and more expensive) than those who didn’t see that image.

This is the power of associations. It is covered in much larger detail in the books Influence and Pre-suasion. I highly recommend reading those and applying what you learn to your daily life. I’m definitely not a professional psychologist so take this post with a grain of salt.

When selling Erlang, whether we are selling it to a customer or trying to convince a developer friend to start using it, we often talk about how Erlang lets you sleep at night, that it is auto healing and always gets fantastic uptimes.

And then we talk about let it crash.

And we describe what it means.

We might as well just say that Erlang crashes a lot and then take the door. It would have the same effect. It doesn’t even stop at programs crashing. You know what else crashes? Cars, planes, trains. Often with disastrous consequences. Is that really the message we want to convey?

They even printed it on a t-shirt! Keep calm and let it crash. It’s the kind of t-shirt you probably shouldn’t wear in an airport, and for good reasons. A few people did, then realized what they were wearing and were not too smug about it.

And yet this is how we sell Erlang.

A better way would be to focus on the positives, of course, but also to make sure that those positives are phrased in a way that prevents bad associations to be formed in people’s minds.

Instead of let it crash, you can say that Erlang has auto healing mechanisms. Healing is a good thing and accurately describes what happens in the system.

Should you need to go into more details, you will probably want to avoid recover from crashes and instead say recover from exceptions. Exceptions are a pretty neutral word and, should you explain what you mean by that, you can talk about exceptions that occur for reasons unrelated to Erlang, like hardware failure or network instability.

The trick is to always use positive words and phrases to describe Erlang, and to use external factors to explain how Erlang deals with failures. Never mention the failures internal to Erlang systems unless you are asked specifically, in which case you can say that the auto healing applies to all exceptions.

The let it crash philosophy is great when learning Erlang or when writing fault-tolerant systems. But it’s not going to convince anyone to use it unless they were already looking for it.

Do you like this post? Tell me on Twitter. I might make more.

Cowboy 2.0 pre-release 4

Tue, 03 Jan 2017 00:00:00 +0100

Cowboy 2.0.0-pre.4 has been released!

This is the new recommended version of Cowboy. While I would not recommend putting it in production just yet, I do recommend you start writing new applications with this Cowboy version.

The most significant changes in the pre-release are:

A new architecture: there now is one process per connection and one process per request. This was done because HTTP/2 allows running requests concurrently.
Stream handlers. Every request, response and data goes through stream handlers. They are meant to replace hooks and more. They will be documented in a future pre-release. Check cowboy_stream and cowboy_stream_h if interested.
Numerous changes to the cowboy_req interface. This is very close to final. Check the manual for what changed.
The Req object is no longer passed in Websocket callbacks.
It is now possible to send frames directly from websocket_init/1.
SPDY support was removed, now that we have HTTP/2.
Update Ranch to 1.3. We still depend on Cowlib master for the time being.
A much improved manual.

I am quite proud of the manual right now. While more improvements can be made, what we have now is way better than before. Feedback for further improvements is welcome!

This is a significant step toward Cowboy 2.0. Almost all the breaking changes are in. A few more pre-releases are planned and will be released on a weekly basis (with exceptions).

Cowboy is now tested and supported with Erlang/OTP 18.0 or above on Arch Linux, FreeBSD, OSX, Ubuntu and Windows 7. Contact me if you can provide permanent access to another platform for the purposes of testing.

Cowboy is now available from four locations:

They are updated at the same time so there is no real difference.

Cowboy 2.0 will be released once all the breaking changes are completed and the temporarily removed features are added back.

Thanks for your patience. I know it took a long time.

Half-price on all donations because I need a new hat:

Ranch 1.3

Mon, 28 Nov 2016 00:00:00 +0100

Ranch 1.3.0 has been released!

This release fixes a number of long standing issues and adds a small number of features:

The ssl application has been added to the list of dependencies. If you don’t need it, you can remove it automatically when fetching Ranch or when building the release. If you do need it, you will no longer have issues shutting down a node because of Ranch.

The ranch:info/0 and ranch:procs/2 can be used to retrieve information about Ranch’s state. Use it for diagnostic and discovery purposes.

SSL listeners can now be configured without a certificate, for setups that make use of the SNI extension.

Transport options are now a blacklist, meaning all unknown options will be accepted. However Dialyzer will warn if said option is not defined in Ranch’s type specifications. Please send a patch when that happens!

Various bugs have been fixed, including the bug where the number of active connections could become negative. Common errors at listener startup should be easier to read (for example when the port is already in use).

See the CHANGELOG for more details.

Ranch is now tested and supported with Erlang/OTP R16B or above on Arch Linux, FreeBSD, OSX, Ubuntu and Windows 7. Contact me if you can provide permanent access to another platform for the purposes of testing.

Ranch is now available from four locations:

They are updated at the same time so there is no real difference.

The most recent Ranch commit is now always signed. You can import the signing key for Loïc Hoguin with:

$ gpg --keyserver hkp://keys.gnupg.net --recv-key 66CCCC8A

The primary key fingerprint is F19F 189C ECC7 4396 99CE DD7A 6EF7 A770 66CC CC8A.

When verifying signatures in git, the following should appear:

gpg: Signature made Sat 26 Nov 2016 12:58:35 PM CET
gpg:                using RSA key 71366FF21851DF03
gpg: Good signature from "Loïc Hoguin <essen@ninenines.eu>" [unknown]
gpg: WARNING: This key is not certified with a trusted signature!
gpg:          There is no indication that the signature belongs to the owner.
Primary key fingerprint: F19F 189C ECC7 4396 99CE  DD7A 6EF7 A770 66CC CC8A
     Subkey fingerprint: FEDA 6E41 B390 F745 A385  5CDC 7136 6FF2 1851 DF03

You can safely ignore the warning if you don’t know what it means, as long as everything else is correct.

Mirrors and signature verification will soon be implemented directly in Erlang.mk. In the meantime, you will need to set them up manually.

Most of this work was done to fix issues in RabbitMQ. Paid customers get priority; contact me if you have some issues that need fixing sooner rather than later.

Expect future releases to be announced in this space.

Thanks for reading!

Mailing list archived

Mon, 29 Aug 2016 00:00:00 +0100

The old mailing list archives have been added to the site, mainly for referencing purposes.

The mailing list has been shut down and all personal information has been deleted.

If you need help with a project, consider either opening a ticket on that project’s issues tracker or going through the community channels (erlang-questions, #ninenines or #erlang on Freenode).

Prefer tickets; often when people have issues it highlights an underlying problem in the project or its documentation.

Thanks.

Website update

Sat, 02 Apr 2016 00:00:00 +0100

Last week-end I updated the Nine Nines website.

I switched to Hugo. The site is now built from Asciidoc documents. You probably saw me switch to Asciidoc for documentation this past year. This is the natural conclusion to that story. The great thing is that with a little bit of Makefile magic I can just copy the documentation files into Hugo and poof, they appear on the website.

I am very happy with that new setup. I can now post my thoughts again. Woo! Expect regular posts from now on. I will try to replace my long series of tweets with posts.

The sections have been rearranged. There used to be a separate training section; now all my services are described in one page. I have also clarified my areas of expertise. There used to be confusion in the past, so now it should be clearer that I am not a distributed systems expert.

On that note, if you are looking for my services right now, I’m not available. I’ll have to work 7 days a week for a while. Try again in a couple months. More on that in a future post.

The documentation becomes a first class citizen. Bullet and Cowlib don’t have proper documentation… yet. I have started working on the Cowlib documentation, and Bullet shouldn’t take too long. All these projects will be documented when Cowboy gets to 2.0, and will all be supported equally. Note that the Cowboy 1.0 documentation still has the old website templates and links. Don’t worry about it.

The mailing lists link has been removed. I did announce a few months back that mailing lists were going to go. They’re still up right now, but not for long. I am planning to put the archives read-only, link to them from a future post and be done with it. If you have a question, open a ticket on Github. Then I can just decide to leave the ticket open if I want to do improvements based on your feedback.

I have replaced most of the "we" by "I". I am a one-man company right now. Have been for a while. Doesn’t make sense to keep a facade. I want to be close to users, not put a barrier between us.

The RSS changed. The old link doesn’t work anymore. The new link is at /index.xml, or /articles/index.xml if you only care about my posts. I guess that’s the one most people want.

I still have some tweaks to do, but it will take a while. My long term plan is to remove Bootstrap, use vanilla CSS and as little JS as possible. The reason for that is that it’s cheaper than upgrading libraries every few years. Life is too short to spend it upgrading JS libraries.

The Erlanger Playbook September 2015 Update

Wed, 02 Sep 2015 00:00:00 +0100

An update to The Erlanger Playbook is now available!

The Erlanger Playbook is a book about software development using Erlang. It currently covers all areas from the conception, design, the writing of code, documentation and tests.

The book is still a work in progress. Future topics will include refactoring, debugging and tracing, benchmarking, releases, community management (for open source projects).

This update fixes a number of things and adds two chapters: IOlists and Erlang building blocks.

Learn more about The Erlanger Playbook!

This is a self-published ebook. The base price is 50€. All proceeds will be used to allow me to work on open source full time.

Thank you for helping me helping you help us all!

Consulting & Training

Wed, 01 Jul 2015 00:00:00 +0100

If you are interested by any of these opportunities, send me an email.

Consulting

You can get me, Loïc Hoguin, author of Cowboy, to help you solve a problem or work on a particular project.

My area of expertise is Erlang; HTTP, Websocket and REST APIs; design and implementation of protocols; and messaging systems.

I can also be helpful with testing or code reviews.

I offer both hourly and daily rates:

200€ hourly rate (remote)
1000€ daily rate (remote and on-site)

For remote consulting, the work can be done by phone, email, IRC, GitHub and/or any other platform for collaborative work.

For on-site consulting, the travel expenses and accomodations are to be paid by the customer. I will also ask for a higher rate if forced to stay on-site for more than a week.

Note that my expertise does not cover all areas where Erlang is used. My help will be limited in the areas of distributed databases, or large distributed systems.

Sponsoring

You can sponsor one of my projects.

Sponsoring gives you:

a direct, private line of communication
the power to make me maintain older versions of my projects (as long as they are sponsoring)
priority when adding features or fixing bugs
advertisement space on this website and in the README file of the project of your choice

Sponsors may choose to benefit from any of these perks.

In exchange sponsors must contribute financially. A minimum of 200€ per month is required. Sponsors may give as much as they want. Payment can be monthly or one-time. Invoices are of course provided.

Erlang beginner training

I would be happy to introduce more people to Erlang. I have a 1-day Erlang training readily available for consumption. The goal of this training is to teach the basics of Erlang systems and programming. It’s a kind of "Getting started" for Erlang.

You can review the training slides.

This training is meant to be given to a large number of people interested in Erlang, as part of a public event, where anyone interested can come.

Another important aspect of this training is that it is meant to be affordable. We want the most people to learn Erlang as possible.

If you have room, think you can gather 20+ people and are interested in sponsoring a training session, then we should talk.

Custom training

I can also provide custom training, tailored to your level and your needs. It can take the form of a class, Q&A or a code review/writing session. I need to know your expectations to prepare an appropriate training.

Custom training rates are the same as consulting rates and the same restrictions apply.

Documentation

Wed, 01 Jul 2015 00:00:00 +0100

Contribute

Do you have examples, tutorials, videos about one or more of my projects? I would happily include them on this page.

Send me an email with the details.

Donate

Wed, 01 Jul 2015 00:00:00 +0100

Like my work? Donate!

Donate to Loïc Hoguin because his work on Cowboy and Erlang.mk is fantastic:

Public talks

Wed, 01 Jul 2015 00:00:00 +0100

Talk requests

Organizing a conference and in need of a speaker for a talk about Erlang and the Web? Need an introduction to Erlang/OTP for your company? Looking for a cool subject for a user group meeting?

Send me an email with the details.

Slogan

Wed, 01 Jul 2015 00:00:00 +0100

Feeling generous? Love reading?
Crowdfund my salary or buy The Erlanger Playbook

The Erlanger Playbook

Thu, 18 Jun 2015 00:00:00 +0100

I am proud to announce the pre-release of The Erlanger Playbook.

The Erlanger Playbook is a book about software development using Erlang. It currently covers all areas from the conception, design, the writing of code, documentation and tests.

The book is still a work in progress. Future topics will include refactoring, debugging and tracing, benchmarking, releases, community management (for open source projects).

The following sections are currently available:

About this book; Changelog; Future additions
Erlang: Building blocks; Patterns
Workflow: Think; Write; Stay productive
Documentation: On documentation; Tutorials; User guide; Manual; README files
Design: RESTful APIs; Lessons learned
Code: Starting a project; Version control; Project structure; Code style; Best practices; Special processes; IOLists; The process dictionary
Tests: On testing; Success typing analysis; Manual testing; Unit testing; Functional testing
Selling Erlang: On persuasion; Don’t let it crash

Read a preview: Special processes

The book is currently 180 pages long. The final version of the book is planned to be between 200 and 250 pages. A print version of the book will be considered once the final version gets released. The printed book is not included in the price.

This is a self-published book. The base price is 50€. All proceeds will be used to allow me to work on open source full time.

You are more than welcome to pay extra by using this second button. It allows you to set the price you want. Make sure to set it to at least 50€ to receive the book.

Make sure to provide a valid email address.

There will be a delay between payment and sending of the book. This process is currently manual.

As the book is a pre-release, feedback is more than welcome. You can send your comments to erlanger@ this website.

The plan is to add content regularly until I run out of things to say. You will receive updates to the book for free as soon as they are available.

Huge thanks for your interest in buying this book!