Circus 0.10 released



Circus LogoI am very happy to announce the release of Circus 0.10.

Circus is a process & socket manager we’ve been developing at Cloud Services to supervise our different processes on our servers It is built using Python and ZeroMQ.

More information is available at

This new version is a major step forward for the project for two reasons. First, we’ve added Python 3 compatibility and second, we did a major refactoring of the core to make it fully asynchronous.

What’s exciting is that most of the work in this release has been done by contributors external to the Cloud Services team, making Circus what we intended at its inception: an open source tool that’s built by and for the Python & Mozilla communities at large.

Below are 3 interviews of Circus contributors. I have asked them the same set of questions.

Fabien Marty

Meet Fabien Marty from Météo France, the French national weather forecast company. He’s one of the main instigators of the core refactoring and explains us what he’s been doing to make it happen.

Tarek: Hello Fabien. Can you tell us who you are?

Fabien: I am Fabien Marty, I am 34 years old and I’ve been working at Météo France as a technical lead for 7 years.

Tarek: How did you end up using Circus?

Fabien: We’re working on a big internal project in my company, where we need to run and supervise a lot of processes (more than 80 per server) and make sure we control every aspect of the processes’ start and stop sequences.

Unlike a classical web stack, we’re working on a stack that receives a huge amount of data — satellite images, numerical modeling data, sensor data, etc. When a server is being stopped, we need to make sure we don’t lose any incoming data. That’s why the stop sequence can last for more than 5 minutes.

We built our own tool to deal with processes, called “launcher”, but it was a bit clunky. Instead of spending more time fixing it, we looked at existing open source projects and found Circus.

Tarek: How did it go with Circus at first?

Fabien: The first tries were very positive. The documentation was clear and detailed and we were able to install it and start using it right away, replacing our own solution.

Tarek: Did you have some issues?

Fabien: Yeah, we had some issues when we tested the stop sequence. As I’ve explained earlier, we have a very specific use case:

we don’t stop our processes with signals – but with flags we’re setting in a Redis server the stop sequence can last for over 5 minutes, because our processes might have to finish processing some data before shutting down we have to start and stop our processes quite often to avoid any memory issues.

Circus was not quite suited for this use case, but we kind of knew we’d hit this problem.

Tarek: What did you do?

Fabien: We started with Alex Marandon (one of the project developers) fixing a few bugs in Circus here and there and pushed them upstream. We also fixed some documentation bugs we found along the way.

Then we wrote a plugin to deal with reloading processes automatically if the process command line is changed in the configuration.

Finally, we worked on the stop sequence: since our processes can take up to 5 minutes to stop and since Circus’ core was not fully asynchronous, this would lock up the event loop and Circus would become unresponsive to any other command during that window.

After some serious design talks with the Cloud Services team, we started to work on a branch to fix this.

Tarek : Was it easy?

Fabien: Not at all ! We knew it was going to be a lot of work. But it was even bigger than we thought. Our first attempt failed: we tried to add a simple callback system hooked into the PyZMQ event loop but that made the code harder to read and understand.

That’s mainly because Circus has high level methods that are doing basic operations, so adding callbacks on those made the whole thing a callback hell.

For our second attempt, we decided to move to a pure Tornado event loop and use its coroutine decorator. That drastically simplified making Circus’ core asynchronous. Moreover, moving Circus code to Tornado coroutines was a no brainer for its PyZMQ compatibility. The library has its own bundled version of a Tornado eventloop but you can use a plain Tornado eventloop if you want, and everything stays compatible.

The bottom line is that we ended up with even simpler code!

Tarek : Are you happy with the result?

Fabien: Yes, very much. There are still a few rough edges,but it works well now and is way better than our initial custom tool. We spent more time than we initially planned but that was worth it – and we don’t regret that investment. :-)

Rémy Hubscher

Meet Rémy Hubscher from Novapost, a French Software company. Rémy is a long time Circus contributor

Tarek: Hello Rémy. Can you tell us who you are?

Rémy: Hey Tarek, I am Rémy HUBSCHER, 26 years old and I work for Novapost as an R&D engineer.

Tarek: How did you end up using Circus?

Rémy : We’re deploying our apps in clusters from 3 to 25 servers, in a private cloud on Amazon and Rackspace. We previously used Supervisor and Gunicorn to deploy them.

Our main motivation for using Circus is its decentralised behavior: every server has its own circus daemon and manages processes there. With a single circus-web dashboard it’s easy to manage them all; we can watch the socket, cpu and memory loads in realtime.

Circus is also easy to configure with Saltstack and it’s dead easy to add new processes and sockets in our stacks.

Tarek: How did it go with Circus at first?

Rémy: Great! Installing Circus in our environment was easy and the clear documentation helped us a lot there.

Tarek: Did you have some issues?

Rémy: Yes, we did have some but my colleague Boris Feld and I contributed fixes to the project that took care of them.

One issue that remains is the fact that it’s not possible to automatically close a socket when all the processes that use it are shut down. But this is going to be fixed soon.

Tarek : What did you contribute in Circus?

Rémy: We started contributing a year ago, by adding simple features like bash autocompletion and the shell in circusctl. We were also involved in brainstorming about the clustering feature, and added automatic UDP discovery of circus daemons.

We also organized a three-day hackaton on Circus in our office to work on the clustering, since we needed it.

Circus-web (the web dashboard) was initially built with Bottle, but we moved it to Tornado for simpler integration with PyZMQ

Tarek : Was it easy?

Rémy: Well, it’s still Python so we found solutions eventually. But it’s important to understand the overall architecture and design, and how Tornado’s async works. But, when there’s an issue, the services team is quite responsive on IRC

Tarek : Are you happy with the result?

Rémy: Very happy. We’ve been using Circus for three apps in production and we’re gradually moving everything else to it.

Tarek : What’s next?

Rémy: We’ll help in reviewing the pull requests on the project and answer any questions – and we will be organizing a new hackaton in early 2014, to tackle more clustering features.

Scott Maxwell

Meet Scott Maxwell, who made Circus python 3 compatible!

Tarek: Hello Scott. Can you tell us who you are?

Scott: My name is Scott Maxwell. I’m a veteran of the video game industry (since 1983) but, recently I have been working on the app store for one of the American car companies.

Tarek: How did you end up using Circus?

Scott: We are currently using Supervisord to run our servers. I used supervisor at Sony Online Entertainment and it worked fairly well for me. We are on Python 3.3 so I had to port supervisor to Py3 about a year ago.

Lately, our restart logic has gotten more complex, since we want to bounce uWSGI without losing any connections. We had to build more and more outside of Supervisor. Also, we started querying Supervisor through the RPC mechanism for monitoring purposes, and it started to crash periodically. The Supervisor team never accepted my changes so Supervisord is still limited to Py2 today and I cannot easily get any fixes they might make.

I discovered Circus through a post on the uWSGI site. When I saw that it was using ZeroMQ and that it supported signals and much richer restart hooks, it seemed like exactly what we were looking for.

Tarek: How did it go with Circus at first?

Scott: The functionality looked great, so we were very excited about the potential.

Tarek: Did you have some issues?

Scott: It was a bit of arough start because of our Py3 requirement. Porting a client/server application to Py3 is much harder because exceptions are caught in one process and sent to the other, losing much of the context. Also, right around the time I was finishing up, the big async upgrade dropped.

Once I got the Py3 port done, I realized that a few other features were missing. For instance, supervisor lets you specify the signal to use for stopping the process. Fortunately this was a very easy feature to add to circus.

Tarek: What did you do?

Scott: I just put my head down and got the initial port done. Once I had basic functionlity in place, I issued a pull request and got everything integrated. From there, the real collaboration began.

Since I was new to the project, I lacked deep understanding of how everything fit together. I spent many hours trying to resolve the resource warnings that the Py3 runtime exposed, without great success. But you, Tarek, were able to fix the majority of them very quickly. When I ran into trouble with the flapping plugin, Alex (@amarandon) jumped straight in. Rémy (@natim) was also very helpful.

Tarek: Was it easy?

Scott: The work was hard, but the collaboration was easy.

Tarek: Are you happy with the result?

Scott: Very happy so far. I expect to have my entire stack moved over to circus on one environment in the next few days. Then I will be able to fully judge the result.

Tarek: What’s next?

Scott: I think Circus is complete for my needs at this point. But it is very comforting to know that if I need a new feature that is general purpose enough, the Circus team will take my change in a timely manner. It gives me great confidence moving forward.

If you are interested in contributing to Circus – visit

Many thanks to Toby Elliott for all the English proofreading. :)

– Tarek, on behalf of the Circus team

Heka 0.4: Dashing through your data

Rob Miller

The Mozila Services team is happy to announce the release of Heka 0.4, the latest version of our logs and metrics processing platform.

We’ve been hard at work fixing bugs and adding features (see the changelog for a full report). One of the most visible change is a complete overhaul of Heka’s internal dashboard UI. Heka’s prior dashboard UI was just a placeholder, the quickest path to exposing the requisite data, but since the previous release we’ve added a much more attractive Backbone.js-based interface with live updating and greatly improved usability. Using the DashboardOutput you’ll be able to see information on how data is flowing through Heka’s pipeline, view time series graphs generated by filters using circular buffers, and examine any other textual data (including JSON, XML, or any other format) that might be generated by a filter. We’ve been using this internally to help make sense of some of our Telemetry data; the attached screenshots show how this looks.

That’s not all we’ve been up to (see the changelog for the full details). Here are more of the highlights:

  • Heka now supports loading and parsing files from a directory (via the LogileDirectoryManagerInput), instead of requiring that all of the files be specified individually. The specified directory will be watched so new folders and files that are added will automatically start being parsed without the need to reconfigure or restart Heka.
  • We’ve added a ProcessInput that will let Heka launch external processes on the host machine, using the process’s stdout output as the input data.
  • The addition of the PayloadJsonDecoder means that you can now map data extracted from arbitrary JSON text to Heka message fields.
  • Sandboxed Lua filters now have access to LPeg (i.e. parsing expression grammar) and JSON decoding libraries, for sophisticated parsing inside your dynamic filter code.
  • The hekad config can now be specified as a directory in addition to a single file, to allow complex configurations to be spread across multiple TOML files.
  • There is now a global working directory configuration option, allowing plugins to store data relative to a root folder rather than having to maintain a full path for each plugin.
  • We’ve greatly improved our input stream parsing, now supporting multi-line records in the input data (with either token or regular expression specified delimiters). That data can come from a log file, an external process, or a TCP, UDP, or AMQP network connection.
  • Similarly, protocol buffer-encoded Heka messages are now supported whether the protobuf stream comes from a file, an external process, or one of the currently supported network protocols (TCP, UDP, AMQP).
  • It is now possible to use sandboxed Lua code in the decoding step, in addition to the filter plugins that have been supported in prior versions.

As always, we love to hear your feedback. Please join us on the Heka mailing list (highly recommended for all Heka users) and in the #heka IRC channel on, and follow the code, submit bugs, and make suggestions on Github.

Heka Health Report
Heka Health Report
Heka Sandboxes
Heka Sandboxes
Sandbox CBuf Output
Sandbox CBuf Output

Introducing the Mozilla Location Service

Hanno Schlichting


The Mozilla Location Service is an experimental pilot project to provide geolocation lookups based on publicly observable cell tower and WiFi access point information. Currently in its early stages, it already provides basic service coverage of select locations thanks to our early adopters and contributors.

A world map showing areas with location data. Map data provided by mapbox / OpenStreetMap.

While many commercial services exist in this space, there’s currently no large public service to provide this crucial part of any mobile ecosystem. Mobile phones with a weak GPS signal and laptops without GPS hardware can use this service to quickly identify their approximate location. Even though the underlying data is based on publicly accessible signals, geolocation data is by its very nature personal and privacy sensitive. Mozilla is committed to improving the privacy aspects for all participants of this service offering.

If you want to help us build our service, you can install our dedicated Android MozStumbler and enjoy competing against others on our leaderboard or choose to contribute anonymously. The service is evolving rapidly, so expect to see a more full featured experience soon. For an overview of the current experience, you can head over to the blog of Soledad Penadés, who wrote a far better introduction than we did.

We welcome any ideas or concerns about this project and would love to hear any feedback or experience you might have. Please contact us either on our dedicated mailing list or come talk to us in our IRC room #geo on Mozilla’s IRC server.

For more information please follow the links on our project page.

Hanno Schlichting, on behalf of the geolocation and cloud services teams

Heka 0.3 released

Rob Miller

Those of us here on Mozilla Service’s Heka team were very pleased by the positive response and interest generated by our initial announcement about the project. And we’re even more pleased by the fact that some of you out there have decided to help out, contributing doc tweaks, bug fixes, and, in some cases, completely new plugins back to the Heka core. All the activity has kept us inspired, and we’ve landed a huge number of fixes and improvements ourselves since then. We’re happy to be rolling these out in a new Heka 0.3 release.

A full list of what’s new in this release can be found in the changelog, but here are some of the bigger features:

  • ElasticSearch output: We had just decided that we wanted to write Heka message data out to ElasticSearch (so we could search through our data using a Kibana dashboard) when we received a pull request from Tanguy Leroux providing exactly that. The screenshot below is of a Kibana dashboard. It is displaying a histogram of the 10 (anonymized) Firefox Sync users who received the most 503 HTTP response codes over a specific period of time, extracted by Heka from our load balancer log files.heka-kibana
  • Restartable plugins: It is now possible to specify any Heka input, filter, or output plugin as restartable, so it will reinitialize itself and start over when encountering an error. This is especially useful for plugins that require persistent connections to external services, as it allows them to reconnect. You can also set them to back off exponentially up to a user-defined cap, or add some timing jitter to prevent several reconnection attempts from happening simultaneously.
  • Resume-from-location log file parsing: When shutting down, LogfileInput will note where it stopped parsing a log file, and will try to pick up from the same location when it restarts.
  • Nagios output: If you use Nagios for monitoring, you can now use the NagiosOutput plugin to generate notifications triggered by Heka messages. Combine this with the ability to do arbitrary data processing in Heka’s dynamic Lua filters, and it becomes very easy to set up ad-hoc notifications for specific targeted events.
  • Improved text parsing: We’ve moved the regular expression match group capturing functionality out of the router and into a decoder, so it won’t slow down routing of messages that don’t use capture groups. We also managed to add some timezone-shifting functionality, for cases where a non-UTC time zone is used but not specified in the timestamps.
  • HTTP input: Thanks to an initial effort by David Delassus, we’ve now got an HttpInput plugin that will make HTTP requests and turn the resulting response bodies into Heka messages. You’ll need a custom Lua filter to parse the results and extract useful data, at least until the helpful decoders that we have under development are ready to take over that job for you.
  • Cloudwatch input & output: We’ve added plugins to get data out of and into Amazon’s Cloudwatch metrics service. They’re not in the Heka core, but they’re in the Mozilla Services repository of custom Heka plugins and are available in the released binaries.
  • New mailing list: There’s a new, dedicated Heka mailing list for announcements about changes to configuration options, Heka behavior, and anything else that might impact running Heka servers. Anyone interested in Heka should check it out!

As you can see, that’s a lot of progress. Big thanks to the Heka team and everyone who sent in patches, bug reports, and suggestions – keep them coming!

Heka is improving rapidly, but it’s still best suited for early adopters at this point. If you’re interested in rolling your sleeves up and digging in, please feel free to check out the binaries, the source code, and the documentation. And don’t forget to join the mailing list, and to drop in to the #heka channel on to ask questions or share your experiences.

Introducing Heka

Rob Miller


We here on the Mozilla Services team are happy to announce our first beta release (v0.2b1) of Heka, a tool for high performance data gathering, analysis, monitoring, and reporting. Heka’s main component is hekad, a lightweight daemon program that can run on nearly any host machine which does the following:

  • Gathers data through reading and parsing log files, monitoring server health, and/or accepting client network connections using any of a wide variety of protocols (syslog, statsd, http, heka, etc.).
  • Converts the acquired data into a standardized internal representation with a consistent metadata envelope to support effective handling and processing by the rest of the Heka system.
  • Evaluates message contents and metadata against a set of routing rules and determines all of the processing filters and external endpoints to which a message should be delivered.
  • Processes message contents in-flight, to perform aggregation, sliding-window event processing and monitoring, extraction of structured data from unstructured data (e.g. parsing log file output text to generate numeric stats data and/or more processing-friendly data structures), and generation of new messages as reporting output.
  • Delivers any received or internally generated message data to an external location. Data might be written to a database, a time series db, a file system, or a network service, including an upstream hekad instance for further processing and/or aggregation.

Heka is written in Go, which has proven well-suited to building a data pipeline that is both flexible and fast; initial testing shows a single hekad instance is capable of receiving and routing over 10 gigabits per second of message data. We’ve also borrowed and extended some great ideas from Logstash and have built Heka as a plugin-based system. Developers can build custom Input, Decoder, Filter (i.e. data-processing), and Output plugins to extend functionality quickly and easily.

All four of the plugin types can be implemented in Go, but managing these plugins requires editing the config file and restarting and, if you’re introducing new plugins, even recompiling the hekad binary. Heka provides another option, however, by allowing for “Sandboxed Filters,” written in Lua instead of Go. They can be added to and removed from a running Heka instance without the need to edit the config or restart the server. Heka also provides some Lua APIs that Sandboxed Filters can use for managing a circular buffer of time-series data, and for generating ad-hoc graph reports (such as the following example) that will show up on Heka’s reporting dashboard:


Heka is a new technology. We’re running it in production in a few places inside Mozilla, but it’s still a bit rough around the edges. Like everything Mozilla produces, however, it’s open source, so we’re releasing early and often to make it available to interested developers (contributors / pull requests welcome!) and early adopters. Here’s a list of resources for those who’d like to learn more:


Implementing cross-origin resource sharing (CORS) for Cornice

Alexis Metaireau

This article is the first technical one on the Mozilla services blog. Expect to read more technical content here in the future. We have been using our respective blogs so far to publish content, but we’ll try to publish the new ones on this blog.

For security reasons, it’s not possible to do cross-domain requests. In other words, if you have a page served from the domain, it will not be possible for it to get data from

Well, it’s possible, using tricks and techniques like JSONP, but that doesn’t work all the time (see the section below). I remember myself doing some simple proxies on my domain server to be able to query other’s API.

Thankfully, there is a nicer way to do this, namely, “Cross Origin Resource-Sharing”, or CORS.

You want an icecream? Go ask your dad first.

If you want to use CORS, you need the API you’re querying to support it; on the server side.

The HTTP server need to answer to the OPTIONS verb, and with the appropriate response headers.

OPTIONS is sent as what the authors of the spec call a “preflight request”; just before doing a request to the API, the User-Agent (the browser most of the time) asks the permission to the resource, with an OPTIONS call.

The server answers, and tell what is available and what isn’t:


  • 1a. The User-Agent, rather than doing the call directly, asks the server, the API, the permission to do the request. It does so with the following headers:
    • Access-Control-Request-Headers, contains the headers the User-Agent want to access.
    • Access-Control-Request-Method contains the method the User-Agent want to access.
  • 1b. The API answers what is authorized:
    • Access-Control-Allow-Origin the origin that’s accepted. Can be * or the domain name.
    • Access-Control-Allow-Methods a list of allowed methods. This can be cached. Note than the request asks permission for one method and the
      server should return a list of accepted methods.
    • Access-Allow-Headers a list of allowed headers, for all of the methods, since this can be cached as well.
    1. The User-Agent can do the “normal” request.

So, if you want to access the /icecream resource, and do a PUT there, you’ll have the following flow:

> OPTIONS /icecream
> Access-Control-Request-Methods = PUT
> Origin:
< Access-Control-Allow-Origin =
< Access-Control-Allow-Methods = PUT,GET,DELETE
200 OK

You can see that we have an Origin Header in the request, as well as a Access-Control-Request-Methods. We’re here asking if we have the right, as, to do a PUT request on /icecream.

And the server tells us that we can do that, as well as GET and DELETE.

I’ll not cover all the details of the CORS specification here, but bear in mind than with CORS, you can control what are the authorized methods, headers, origins, and if the client is allowed to send authentication information or not.

A word about security

CORS is not an answer for every cross-domain call you want to do, because you need to control the service you want to call. For instance, if you want to build a feed reader and access the feeds on different domains, you can be pretty much sure that the servers will not implement CORS, so you’ll need to write a proxy yourself, to provide this.

Secondly, if misunderstood, CORS can be insecure, and cause problems. Because the rules apply when a client wants to do a request to a server, you need to be extra careful about who you’re authorizing.

An incorrectly secured CORS server can be accessed by a malicious client very easily, bypassing network security. For instance, if you host a server on an intranet
that is only available from behind a VPN but accepts every cross-origin call. A bad guy can inject javascript into the browser of a user who has access to your protected server and make calls to your service, which is probably not what you want.

How this is different from JSONP?

You may know the JSONP protocol. JSONP allows cross origin, but for a particular use case, and does have some drawbacks (for instance, it’s not possible to do DELETEs or PUTs with JSONP).

JSONP exploits the fact that it is possible to get information from another domain when you are asking for javascript code, using the <script> element.

Exploiting the open policy for <script> elements, some pages use them to retrieve JavaScript code that operates on dynamically generated JSON-formatted data from other origins. This usage pattern is known as JSONP. Requests for JSONP retrieve not JSON, but arbitrary JavaScript code. They are evaluated by the JavaScript interpreter, not parsed by a JSON parser.

Using CORS in Cornice

Okay, things are hopefully clearer about CORS, let’s see how we implemented it on the server-side.

Cornice is a toolkit that lets you define resources in python and takes care of the heavy lifting for you, so I wanted it to take care of the CORS support as well.

In Cornice, you define a service like this:

from cornice import Service

foobar = Service(name="foobar", path="/foobar")

# and then you do something with it
def get_foobar(request):
    # do something with the request.

To add CORS support to this resource, you can go this way, with the cors_origins parameter:

foobar = Service(name='foobar', path='/foobar', cors_origins=('*',))

Ta-da! You have enabled CORS for your service. Be aware that you’re authorizing anyone to query your server, that may not be what you want.

Of course, you can specify a list of origins you trust, and you don’t need to stick with *, which means “authorize everyone”.


You can define the headers you want to expose for the service:

foobar = Service(name='foobar', path='/foobar', cors_origins=('*',))

@foobar.get(cors_headers=('X-My-Header', 'Content-Type'))
def get_foobars_please(request):
    return "some foobar for you"

I’ve done some testing and it wasn’t working on Chrome because I wasn’t handling the headers the right way (The missing one was Content-Type, that Chrome was asking for). With my first version of the implementation, I needed the service implementers to explicitely list all the headers that should be exposed. While this improves security, it can be frustrating while developing.

So I introduced an expose_all_headers flag, which is set to True by default, if the service supports CORS.

Cookies / Credentials

By default, the requests you do to your API endpoint don’t include the credential information for security reasons. If you really want to do that, you need to enable it using the cors_credentials parameter. You can activate this one on a per-service basis or on a per-method basis.


When you do a preflight request, the information returned by the server can be cached by the User-Agent so that it’s not redone before each actual call.

The caching period is defined by the server, using the Access-Control-Max-Age header. You can configure this timing using the cors_max_age parameter.

Simplifying the API

We have cors_headers, cors_enabled, cors_origins, cors_credentials, cors_max_age, cors_expose_all_headers … a fair number of parameters. If you want to have a specific CORS-policy for your services, that can be a bit tedious to pass these to your services all the time.

I introduced another way to pass the CORS policy, so you can do something like that:

policy = dict(enabled=False,
              headers=('X-My-Header', 'Content-Type'),

foobar = Service(name='foobar', path='/foobar', cors_policy=policy)

Comparison with other implementations

I was curious to have a look at other implementations of CORS, in django for instance, and I found a gist about it.

Basically, this adds a middleware that adds the “rights” headers to the answer, depending on the request.

While this approach works, it’s not implementing the specification completely. You need to add support for all the resources at once.

We can think about a nice way to implement this specifying a definition of what’s supposed to be exposed via CORS and what shouldn’t directly in your settings. In my opinion, CORS support should be handled at the service definition level, except for the list of authorized hosts. Otherwise, you don’t know exactly what’s going on when you look at the definition of the service.


There are a number of good resources that can be useful to you if you want to either understand how CORS works, or if you want to implement it yourself.

Of course, the W3C specification is the best resource to rely on. This specification isn’t hard to read, so you may want to go through it. Especially the “resource processing model” section

Finally, you may want to have a look at the actual implementation in Cornice.

Retiring Firefox Home


From the early days, Mozilla has been focused on empowering users across platforms and devices.  We released Firefox Home as an experiment in bringing a part of the Firefox experience to iOS, focusing on Firefox Sync.  This project provided valuable insight and experience with the platform, but we have decided to remove Firefox Home from the Apple App Store and focus our resources on other projects.

For those interested in continuing to use or improve the iOS Sync client that Firefox Home is built on, we have made the source available on GitHub, free of Mozilla trademarks and ready for independent development.  As with all Mozilla projects, we ask developers to be aware of the Mozilla trademark policy.

We remain committed to providing compelling user experiences across as many platforms and devices as possible and will continue to explore the best ways to provide great experiences to iOS users.

– mconnor, on behalf of the Firefox and Services teams

Add-on Sync Coming to Firefox

Gregory Szorc


We strive to make your online experience better and we have a new feature in the latest Firebox Beta we think you’ll love: add-on sync.

Add-on sync does what its name implies: it synchronizes add-ons between profiles connected with Firefox Sync. Specifically, it will install, uninstall, enable, and disable add-ons across your devices as you do.

If you are a new Sync user, add-on sync will be enabled by default. However, since Mozilla cares about your privacy and we don’t want to do anything without your explicit permission, existing Sync users will need to manually opt in to the feature. This can be done through the Sync tab in Firefox’s Preferences window. Explicit instructions are available at

Once you have add-on sync enabled, you don’t need to do anything special to get your add-ons to sync. As you use your browser, Sync will run in the background. As it does, the current state of your add-ons will be collected and sent to the sync server. As you use your other devices, Sync will apply changes to your local Firefox. Add-on sync runs in the background and you won’t see any pop-ups indicating it is running. And, since some add-ons require a restart for changes to be made, you may not see your new add-ons until you restart your browser. Our studies show that over 99% of Firefox users restart their browser at least daily, so you shouldn’t have to wait too long. If you are curious, you can open the Add-on Manager (about:addons) and see what changes will occur on the next restart.

There are a number of challenges involved with synchronizing add-ons. Because of this, the scope of the initial add-on sync feature has intentionally been limited. For this initial release, an add-on will be synchronized only if all of the following criteria are met:

  • It is an extension or theme
  • It is installed from
  • It is publicly listed on the add-ons site
  • It is installed into the current profile by the user

For now, add-ons are only synchronized between identical application types – changes to a desktop browser will only affect other desktop browsers and changes to a mobile browser will only affect other mobile browsers. Greater functionality between desktop and mobile will come in the future.

Security and privacy are important concerns when designing add-on sync. As with all your Firefox Sync data, add-on data is encrypted in your browser before being transmitted to the Sync server. So, people in the cloud can’t tell what add-ons you have installed, even if they wanted to know.

Add-on sync is a feature in progress and are working on further updates to include in later Firefox releases. You can learn more about add-on sync including background on some key design decisions at If you would like to get involved, instructions for reaching us can be found at

We hope add-on sync makes managing your online experience with Firefox a little easier. Happy syncing!

Focusing on the integrated Firefox Sync experience



Today we are announcing the end of official support for the Firefox Sync add-on, focusing resources on improving and supporting the Sync experience for those users in the latest versions of Firefox. This means that the add-on will not be available at and the only way to get Firefox Sync features will be to upgrade to the latest version of Firefox where they are built into the browser.

A lot has happened since Mozilla launched Weave back in 2007: We launched Weave as an add-on, rebranded it as Firefox Sync, moved Firefox to Rapid Release and eventually integrated Firefox Sync into the browser. Our goal has always been to offer the best service so you can seamlessly share your Firefox experience across devices

Since the launch of Firefox 4, the browser comes bundled with all the goodness of Firefox Sync. Because of the continued support for Firefox 3.6, we have continued to support the add-on. This causes a great deal of engineering & testing headaches between the add-on and the integrated Sync. We also see a lot of users unhappiness since the add-on does not support some of the newer, highly desired, features we’ve added to Sync.

Keep in mind the add-on will still be usable for now, but we will not guarantee its reliability in the following months and foreseeable future. We ask everyone still using the Sync add-on to move to newer versions of Firefox to enjoy the best online and Sync experience.

— mconnor & ibai, on behalf of the Services and SUMO teams

Introducing Project Sagrada


Project Sagrada is a Services project to build a solid and stable platform to develop new and highly scalable services and applications to support Mozilla’s mission. The initial goal is to make it easy for internal developers to build reliable and high performance backend services, easily deploy them to test or production, and have their deployed services be robust and easy to manage. Based on the current product direction for the known services, this set of capabilities are both the hardest to get right at scale and the most important shared aspects between all of the services.

The first phase of the project will focus on adding capabilities (abuse detection, metrics, crypto, and more) and tooling (testing support, build automation, and packaging/dependency tracking) to the current python services-core framework. Later phases will provide better support for building and deploying web-based applications, increasing amounts of “self-serve” capabilities around deployment and provisioning, and options for isolating experimental/external apps.

Creating a core set of stable and reusable code libraries and, where appropriate, web APIs alongside the programmatic ones, technology choices (i.e. which key-value store, which auth technology, which abuse detection system, etc) are abstracted away and allow the user to focus on the core of code that is unique to their app. We believe this is the best and fastest path to a solid, scalable and highly capable base platform that can be used as a framework and as individual components.

How We Get There

Our efforts to date have focused on building out and refining capabilities as they arisen on a project by project basis. While still maintaining a focus on products we support and products we are working towards launching, our priorities in Q4 and beyond will shift towards building generic support for applications that teams across Mozilla will develop.
We recognize that this is a highly ambitious project, and can be effectively described as completely open-ended. While true, there is much to be said for where we can get in finite time along that path.
The structure of a service generally follows the same pattern: a set of API entrypoints, some authentication, controller logic, storage, and a response. Then there’s the additional questions of deployment, monitoring and maintenance. All of these things combine to make writing an app from scratch a challenge, especially if you want to build it to scale out of the box.

What if we could make that 20% easier? That would seem to be a big win, both internally and externally. We probably have that today – use our base application and generating your entrypoints is pretty simple. Sure, you still have to write all your controller logic, and figure out your storage story, etc, but hey, 20% is a pretty good start.

What if we could make it 40% easier? Simple key-value storage APIs and a strong metrics component might do that. And so on and so on.

Why not use one of the many available frameworks out there? We could probably pick up a quick win building on one of them, and we haven’t ruled out the possibility. However, most frameworks are built to support HTML generation, and while they’re capable of creating the appropriate responses, they pull in a lot of additional functionality that we do not need or use, and if it gets in the way of our ability to scale, it’s a nonstarter. Making it easy for us to get to 30% is a great idea, as long as it doesn’t make it much harder to get to 50%.

The ultimate goal is 100%. This goal is impossible, as any computer programmer can tell you, and trying to achieve that produces the infinite horizon that makes this work seem so vast. But setting that as the goal misses the point – getting a user 75% of the way to having a stable, scalable backend application that does what they need would be an enormous game-changer in the application creation space.

The Roadmap

We’re still in the bootstrapping phase of the project, so we’re building some of the obvious pieces, and gathering feedback on the rest. Please see the wiki for the current set of features we’re looking to add to the platform.

— mconnor & telliott, on behalf of the Services team