Planet Jabber

Distribuir contenido
Planet Jabber -
Actualizado: hace 1 semana 5 días

ProcessOne: TextOne 4.0.0 released

28 July, 2014 - 14:16

We just released a new version of TextOne iOS messaging client.

This is a version improving compliance with iOS 7 (and introduce a new and lighter design). It also improves networking stack and time to connect / retrieve messages.

It also fixes a long standing crash issue on launch that could sometime happen.

It can be downloaded from Apple App Store: TextOne on Apple AppStore.

ProcessOne: ejabberd Community 14.07

23 July, 2014 - 10:40

Two months after the previous huge release, we are keeping the pace with a new feature packed version. We added major improvements regarding to SIP support, Carbon copy, stream management and added Riak as a new supported database.

ejabberd Community 14.07 includes several improvements and many bugfixes over the previous 14.05 release:


  • SIP Outbound (RFC 5626) support
  • New option always_record_route
  • New options record_route and routes

Carbon Copy (XEP-0280)

  • Don’t log MUC messages with hint
  • Don’t carbon copy messages with hint
  • Let is_carbon_copy/1 recognize carbons
  • Don’t send XEP-0280 v1 copies back to sender

Stream Management (XEP-0198)

  • XEP-0198: Terminate session if stanza queue becomes too large
  • XEP-0198: Don’t exit on socket send failure
  • XEP-0198: Don’t drop session on failed resume
  • XEP-0198: Check whether routed packets are stanzas


  • Riak support

Install and config

  • New options log_rotate_count, log_rotate_size and log_rate_limit
  • html guide is now generated when building source tarball
  • Use p1_utils, and move treap.erl to p1_utils
  • Get rid of p1_mnesia file
  • old release notes are not installed anymore
  • Don’t “forget” listener options
  • Always enable STUN at compile time
  • Do not check for Erlang apps at configure time
  • Add –enable-riak configure flag


  • Add tests for stream management
  • Add tests for mod_carboncopy
  • Add tests for mod_caps
  • Improve MUC test cases
  • Travis CI: Enable Riak tests
  • Add Riak backend to the testing suit
  • The test suite no longer fails without –enable-transient_supervisors.


  • MUC messages with ~ were not logged (EJAB-1696).

And many bugfixes all over the source code.

This release requires at least Erlang/OTP R15, and works perfectly with R16B03, and will work with R17 if Riak support is disabled (–disable-riak).

When building from Git, “make install” now fails if you do not run “make doc” before due to doc/guide.html no longer being in the repo. If you don’t have necessary tools to build the guide, just pick the one from ejabberd sources tarball.

As usual, the release is tagged in the Git source code repository on github

The source package and binary installers are available at ProcessOne

If you suspect you found a bug, search or fill a bug report in Jira

Fanout Blog: Mongrel2 HTTP server now in Debian/Ubuntu

18 July, 2014 - 19:58

Mongrel2 is a fast and simple HTTP & WebSocket server that communicates to backend workers via ZeroMQ. It does one thing and does it very well, making it an ideal part of a componentized architecture. The code is event-driven, allowing it to support thousands of concurrent connections and also asynchronous behaviors. These properties are especially important to realtime applications.

Fanout has been one of the most active contributors to the Mongrel2 project over the past year, adding features such as TLS SNI and improved streaming capability. We've also been working on making the server easier for people to get started with. And with that, we are proud to announce official packages for Debian and Ubuntu!


ProcessOne: GigaOM releases report on brand engagement through in-app communication

17 July, 2014 - 08:22

Mobile apps now allow businesses to be more responsive to users than ever before, but this interaction has created new customer expectations. While a mobile app must fulfill its core function, it must also continue a dialog with its users to remain relevant. Doing so requires ongoing, intelligent, targeted outreach to customers and an extension of customer-service strategy into the app itself.

Brands and businesses that develop mobile applications must be aware of the demands and limits of an increasingly sophisticated mobile audience, build a communications strategy that spans the appropriate communications channels, and play to the strength of each channel.

In-App communication gathers as a whole three mobile oriented channels:

  • Native Push Notification services.
  • In-app notifications.
  • In-app chat.

Together, they form the three main components of brand realtime user relationship management on mobile. Today, those three approaches are your best bet to establish a strong communication channel that provides a real value to your users.

The field is quite new and we are glad to be able to share with GigaOM the first research on brand engagement through in-app communication.

The document gives examples, insights and best practices to tighten the links with your mobile application user base and provide them the best value at the right time.

We are pleased to make this report available for free to people interested in Boxcar Push Notifications Service. Join our mailing list focused on leveraging in-app communication for brands to receive report download link »

(Do not worry, it’s low traffic, no spam, we only care about the value we give you).

Report is also available to GigaOM subscribers.

Isode: New “Military forms using XMPP” whitepaper and access to our Military forms Demo

7 July, 2014 - 14:11

Forms are important for military operations, and there is often a need to handle forms quickly and share with a large number of users, such as Medical Evacuation (MEDEVAC) alerts.

XMPP based open standard instant messaging is widely used by military organizations and is a sensible framework for sharing forms. Our new whitepaper [Military Forms using XMPP], published on the Isode website today, looks at the requirements for military forms and how the XEP-0346 “Forms Display and Publishing”(FDP) can be used to provide real time military forms. It looks at how capabilities provided by M-Link support military forms using FDP, and how gateways can enable integration with other services. FDP is supported in the most recent R16.2 release along with FDP Management in M-Link Console.

We have created a military forms demonstration web site using FDP and Isode’s demonstration Web FDP client. To use this demonstration you will require information on the demonstration accounts. Please contact for access to this information.

If you wish to set up an FDP system, this is explained in our FDP Evaluation Guide which will show you how to set up a basic XMPP Form Display and Publishing (FDP) configuration using Isode’s M-Link server, Isode’s Demonstration web client and Isode’s FDP Demonstration Desktop Client.

Peter Saint-Andre: Thoreau and the Seasons of Man

5 July, 2014 - 00:00
This evening I finished reading Thoreau's first book, A Week on the Concord and Merrimack Rivers. Toward the very end, he writes as follows:

Fanout Blog: You might not need a WebSocket

1 July, 2014 - 01:32

Before I begin, I want to say that WebSockets are great. I've even implemented RFC 6455 myself in Zurl and Pushpin, which are used by the service. also supports WebSockets via its XMPP-FTW interface using Primus.

However, after spending quite some time working on large distributed applications and gaining a greater appreciation of REST and messaging patterns, I feel that much of what typical web applications want to accomplish with WebSockets (or with socket-like abstractions) is perhaps better solved by other means.


ProcessOne: Cardinality Estimation

30 June, 2014 - 08:47

In server world, we always need to maintain some metrics; We need to measure to improve. A very common one being “unique active user” per unit of time. While this is really easy to describe, it’s complex when it comes to implementation.

Naive implementation logs all events (let say, user connection), either on memory or disk, and count number of unique entries in the log on a given time frame by removing duplicates.
Well, this is really consuming and can not scale on real server handling millions of user. Better take the probabilistic estimation way… and here comes a very impressive algorithm.

Back in 2007, a team at INRIA published a paper about an efficient algorithm for estimating the number of distinct elements, known as the cardinality, of large data ensembles. This paper by Philippe Flajolet, Éric Fusy, Olivier Gandouet and Frédéric Meunier can be found here.
There are many blogposts about it already, but this paper is worth reading to make your day !

The HyperLogLog algorithm is just genius thanks to uncommon approach that make it require only few Kilobytes of data to give an accurate cardinality estimation on large sets. Basically, the idea is to rely on probability of distribution of user id numeric representation (hashes).

The HyperLogLog algorithm works for maximal cardinalities in the range [0..109] while handling a number of registers (m) in the range [24..216].
Of course, using less registers reduce computing time and memory use, at cost of cardinality estimation precision.

So, what is best value of m (number of registers) for my need ?
Well, that all depends ! Of course, we will run it on huge production services, so it must be fast. It also must be accurate in most cases.

Let’s assess errors margin and approximations

From the original paper, we have an estimation of the standard error of the algorithm:

Let σ ≈ 1.04/√m represent the standard error; the estimates provided by HYPERLOGLOG
are expected to be within σ, 2σ, 3σ of the exact count in respectively 65%, 95%, 99%
of all the cases.

bits m σ 2σ 3σ 10 1024 ±3.25% ±6.50% ±9.75% 11 2048 ±2.30% ±4.60% ±6.90% 12 4096 ±1.62% ±3.26% ±4.89% 13 8192 ±1.15% ±2.30% ±3.45% 14 16384 ±0.81% ±1.62% ±2.43% 15 32768 ±0.57% ±1.14% ±1.71% 16 65536 ±0.40% ±0.81% ±1.23%

Let say one need cardinality approximation ±~3%, with a trust level of 95%.
One can get cardinality ±3.26% using only 4096 registers.
Now if i want error <2% with a trust level of 99%, I must use 215 or 216 registers.

Now, what about execution runtime and memory consumption ?

Let’s play with the erlang implementation available on github thanks to Shyun Yeoh (vaxelfel).

First, let’s check how much Erlang VM memory is needed to store one hyperloglog record, given number of registers m

m memory in bytes 2048 2845 4096 6173 8192 13341 16384 28701 32768 62469 65536 131101

The memory consumption is approximative, and depends on default heap size and other low level memory parameters. But it gives the big picture anyway.

At last, let’s try to update an HyperLogLog record one million time with unique values, and report time and error margin:

m error ms 2048 +3.41% 2178 4096 +2.31% 2414 8192 +0.71% 6343 16384 -0.49% 11100 32768 +0.16% 25667 65536 +0.10% 9584

Without surprise, the consumed time in HyperLogLog update is almost proportional to data structure size in memory, at one exception.
When using m=216, there is an approximate 5x speedup on this implementation. I guess this is related to some binary matching optimization on Erlang VM but I could not get measures clearly stating this.


A production server never run on its own. The magic with HyperLogLog algorithm is that merging data from different servers is possible as long at they run with the same hash function and the same number of registers.
The resulting estimate of merged data gives cardinality for whole cluster. Bravo !


For rapid cardinality estimation using Erlang implementation, there is two main use of HyperLogLog:

  • For fast operation and minor resources consumption, use of 212 registers.
  • For precise estimation, if memory is a problem, use 214 registers, if CPU is a problem use 216 registers.

Alexander Gnauck: BOSH prebind in MatriX and attach to Strophe

29 June, 2014 - 10:13
Why prebind and attach?

When your web application is already logged in and needs to establish a XMPP connection you have a seamless and secure solution with prebind and attach. You don’t have to ask the user to provide the credentials again, and you don’t have to render the credentials insecure in your HTML output.

Prebind with MatriX

This code shows how to prebind a session with MatriX. On success MatriX raises the OnPrebind event and provides the Jid, Rid and Sid in the PrebindEventArgs. The Rid is the last one used by MatriX and must be incremented before getting passed to Strophe.
You can render this params in a new page and attach the session with Strophe.

var xmppClient = new XmppClient { Username = "testuser", XmppDomain = "", Password = "secret", Transport = Matrix.Net.Transport.BOSH, Uri = new System.Uri("") }; // subscribe to the prebind event xmppClient.OnPrebind += client_OnPrebind; void client_OnPrebind(object sender, PrebindEventArgs e) { prebindArgs = e; are.Set(); } // start prebinding with MatriX now xmppClient.Prebind();

You can download a complete ASP .NET example project here:
Bosh Prebind Example

Alexander Gnauck: Await MatriX

29 June, 2014 - 10:12

C# 5.0 brought us the great new Async and Await feature. It helps to avoid bottlenecks when it comes to performance or to improve the responsiveness of UI applications.

Both of them were never a problem in MatriX, because its completely asynchronous internal. We helped GUI developers a lot by handling all multi threading and invoking to GUI threads inside of MatriX.

But the traditional techniques for writing asynchronous code based on callbacks can be complicated, difficult to maintain, hard to debug, error prone, and often ends in the callback hell.

Here is an for requesting a vcard with the new async await API

var viq = new VcardIq { To = "", Type = IqType.get }; var vcardResult = await xmppClient.IqFilter.SendIqAsync(viq);

Here is a more complex example of creating a MUC room containing the following steps.

  • request our reserved nickname
  • create the room with the requested nickname
  • request the room configuration
  • submit the room configuration

With the old async callback based API this simple task requires you to setup multiple callbacks or listen to multiple events. This splits your code in many different functions, or you use many inline lambdas. Error handling gets complicated and the control flow is hard to follow.

So look at the code with the new async await API. It looks like simple synchronous code. Easy to read with simple try/catch error handling, fully asynchronous. The compiler is doing all the complicated work for us.

try { Jid roomJid = ""; var nick = await mucManager.DiscoverReservedNicknameAsync(roomJid); var createResult = await mucManager.CreateRoomAsync(roomJid, nick); // check from createResult if the room requires configuration if (createResult.Error != null && createResult.Error.Condition == ErrorCondition.ItemNotFound) { var roomConfig = await mucManager.RequestRoomConfigurationAsync(roomJid); var data = ProcessRoomConfig(roomConfig); var submitResult = await mucManager.SubmitRoomConfigurationAsync(roomJid, data); } } catch (TimeoutException ex) { // handle exceptions here } catch (Exception ex) { // handle exceptions here }

Many of our customers are tied to older .NET versions and not able to use the latest and greatest features. This is the reason why we will offer separate builds for .NET 4.5 with the async API extensions, and .NET 3.5 builds.

We will upload new .NET 4.5 builds soon, so stay tuned. If you can’t await then contact us directly ;-)

Alexander Gnauck: NuGet package available

29 June, 2014 - 10:12

Some of our customers asked for NuGet packages. The roadmap of our development is heavily impacted by our customer requests. We are always receptive to your criticism, ideas and praise

When you prefer NuGet for managing your references then you can start using it for MatriX as well now.

You can find the MatriX NuGet package here:

Alexander Gnauck: MatriX Screencast #2 published

29 June, 2014 - 10:12

The 2nd screencast has been published. This tutorial covers the following:

  • installation of a local Openfire XMPP server
  • configuration of the server
  • creating user accounts
  • starting XMPP client development on the local server

More screencasts and tutorials for other XMPP server software is coming.

Alexander Gnauck: Task-based asynchronous MUC room configuration

29 June, 2014 - 10:12

The question on how to change the configuration of a Muc room, or how to make a room persistend comes up very often in our forums and support enquiries.

Here is a small example using MatriX and the task based asynchronous pattern.

In MatriX most of the Muc functionality can be accessed with the MucManager component. We assume that you are familiar with the MucManager and have joined or created Muc rooms already.

In this small example we want to make an existing room persistent.
Four steps are required to achieve this:

  1. Request the current room configuration.
  2. Parse the current room configuration in code or render it to a GUI.
  3. Create the xdata result, either in code or from the user input in the GUI.
  4. Submit the new room configuration.

Here is the code:

private async void MakeRoomPersitent(Jid roomJid) { // Step 1, request the room configuration var iq = await mm.RequestRoomConfigurationAsync(roomJid); if (iq.Type == IqType.result) // only proceed on result { // Step 2 and 3, parsing the current config and // creating the result is done in the same loop here. var xdata = iq.Query.Element<Data>(); var xDataResult = new Data { Type = FormType.submit }; foreach (var field in xdata.GetFields()) { var retField = new Field() { Type = field.Type, // keep the type Var = field.Var // keep the var }; // we are changing the muc#roomconfig_persistentroom only // other fields get copied only with the existing values if (field.Var == "muc#roomconfig_persistentroom") retField.AddValue(true); else retField.AddValues(field.GetValues().ToArray()); xDataResult.AddField(retField); } // Step 4, submit the changed configuration back to the server (room) var submitIq = await mm.SubmitRoomConfigurationAsync(roomJid, xDataResult); if (submitIq.Type == IqType.result) Debug.WriteLine("success"); else Debug.WriteLine("something went wrong"); } }

If you want to render Data froms to UI, and generate responses from the UI input, you can use the sources in the MatriX.Ui project which can we downloaded from here:

Alexander Gnauck: MatriX Screencast #1 published

29 June, 2014 - 09:59

We have published the first MatriX screencast. This small tutorial shows you howto download and install the MatriX XMPP SDK. And howto login with the MiniClient example using your Gmail Id.

You an find the screencasts here. More screencasts are coming soon…

ProcessOne: Google I/O: A couple of days with an Android Watch

27 June, 2014 - 22:53

I was at Google I/O conference during the past week. I will not yet comment on the overall tone of the conference, as I need to give it more thought.

However, I can comment on the Smartwatch Google gave to Google I/O attendees. I picked LG Watch and gave a try over two days on Android Wear. I picked this one over the other option, Samsung Gear Live, because I was told it would not be very compliant with a non-Samsung device (I use a Nexus 5).

The software is somewhat nice. This is basically a quick access to Google Now: simply making it more accessible may increase its usefulness. Integration with an Android phone is ok. The watch is especially nice for notifications and I was thrilled seeing BBC Sport World Cup goal alerts on my wrist. Application do not have to make any special change to display their notifications on the watch.

However, there is a big catch. Wearing the device hurts. Wearing the LG strap is really painful for my wrist, and various attempt at adjusting it did not help. After wearing it for a while, all I want is removing it to ease the pain. That’s what I did now. I just got back to using my phone as before, by watching its screen when I need to.
I (almost) gave up after a couple of days.

If Google (Or Apple) expects to be successful in wearable, they need to focus on a perfect, flawless user experience. The device itself should be pleasure and the added value must be obvious.

Users expectation are going to be very high. From watches to glasses, in wearable computing, “good enough” is simply not enough.
… And it looks that I am not the only one to think so. You can read excellent Ben Thompson article for more: Android where?

Isode: Isode support to Boeing and NCI Agency at Unified Vision 2014

16 June, 2014 - 14:49

Over the last couple of years we’ve been conducting both ground and flight trials with a number of military aircraft operators to look at addressing the problems of text chat over constrained links (high-latency, unreliable connection, low-bandwidth).

Text chat has become a vital capability for the modern warfighter but most modern text chat deployments have significant problems, both architectural and functional, in the constrained link environment.

Addressing these problems has been a high priority for our development team and we believe that our M-Link XMPP server product now leads the field in this environment.

We continue to participate in trials whenever we’re given the opportunity, which is why we were very happy to support Boeing and NATO’s NCI Agency in the recent Unified Vision 2014 exercise, the largest ever test of NATO’s intelligence, surveillance and reconnaissance (ISR) capabilities.

M-Link capabilities, including Federated Multi-User Chat and submission of Tactical Reports (TACREPS) using dynamic chat forms, were extensively tested over a 10 day period. We’re very happy with the feedback and results we got from the tests, which will enable us to make even more improvements to M-Link’s performance.

The results from Unified Vision will be used as the baseline for implementation of a Joint ISR Initial Operational Capability, in 2016, for the NATO Response Force.

Peter Saint-Andre: Cultivating Your Higher Ground

14 June, 2014 - 00:00
I've been reading Thoreau's letters to Harrison Blake, which are a veritable mine of philosophical insights. For the purposes of writing Walking With Thoreau, I'm especially interested so far in a fascinating vision he draws of cultivating the spiritual reaches of life (letter of May 28th, 1850):

ProcessOne: A week at Apple WWDC – early thoughts

11 June, 2014 - 08:56

I have spent the past week in San Francisco at Apple WWDC (Worldwide Developers Conference). It was the richest and more energetic WWDC I have ever participated in. Apple announced a lot of new features that are going to benefit our XMPP and push platforms for iOS and our mobile software.

But, as I still ponder the overall implications of the conference, I wanted to share some of my initial thoughts and debunk two misunderstandings.

The general media failed to understand how major this conference was

My first comments are about the gap in understanding by the general media regarding this conference. In France, for example, Le Monde, one of the major French newspapers, wrote that Apple is refreshing its iOS, but no revolution has been announced. This was a common pattern found in the non-technical press.

This fails to measure how radical the change in iOS and OSX was for the development community. Apple did open a new set of API, allowing interapplication communication and custom widgets for the notification center. It also created a new language to improve the overall platform for developers. More generally, most of the 100+ talks at WWDC were about granting more power to developers on various levels, improving API, and solving common and old technical limitations. Example from a random conference: Core data performance has been improved on two use cases—batch update and async fetch. This announcement alone is a reply to a complaint on core data performance for mass record changes; see On switching away from Core Data by Brent Simmons.

At the conference, there was general agreement that nearly all the wishes and complaints from Apple developers had been satisfied. Some said it was the most important Apple developer event in 10 years, with beta testing, performance, more powerful API, and even a new modern language called “Swift”. In a single week, the Apple development landscape radically changed. Marco Arment, a high-profile developer, noted that Apple opened new territory. Casey Liss said that during this conference, Apple changed its mindset, and its new message was about building a platform together with the development community.

WWDC is a developer conference, and it is getting back to its roots. The media was expecting hardware or software announcements. Apple has announced new hardware in the past at WWDC (examples are iPhone 3GS or iPhone 4) but was disappointed by announcements targeted at the developer community. However, it happens that the developers are the ones building the ecosystem. What Apple gives them is the tools that will make the platform much better and more powerful in the coming years. It will take a while to have everyone notice, but it is a profound change. This is a seed to improve the application ecosystem on iOS for the coming years.

“Swift” impact

Apple surprised the public with an announcement of a brand-new programming language. Despite it having being in development for four years, Apple managed to keep the secret until now. Launching a new programming language will have a huge impact on the development community. For developers, it was like Christmas in June.

That said, the impact was, again, largely misunderstood. It was described as a way to make the development with iOS simpler and accessible to more developers. However, having assisted at many talks on Swift, it is clear to me that the goal is not to appeal to a mass of new developers. Google is using Java—one of the most taught languages in the world—on this platform, and there is no way to compete with the ubiquitous knowledge of that platform among developers. Code factories are mostly based on Java skills, and Java is a programming language with a massive amount of manpower behind it.

What Apple proposes with Swift is that it be a language that is efficient, making it very adequate for mobile environments. It also is a language that is much more expressive and enjoyable to write than Objective C, the existing de facto language for iOS and OSX programming.

What Swift is not is a simple language. It is extremely expressive and powerful—and, as such, it requires a deep understanding of many programming paradigms (object oriented, functional). Developing for iOS and OSX requires knowledge of the many features available in the frameworks. No matter how you express your code, learning those frameworks takes time. You will still need a great deal of skills to write mobile apps.

Swift is there because Apple wants to help attract good developers—not simply a mass of developers. Apple wants to appeal to developers that are constantly looking for a better way to express their code and improve the performance and maintainability of their software.

Reading the first articles and analysis on Swift is by no means about “writing iOS or OSX software in 21 days”. It is profound how the articles analyze the influence and semantics of the language, and what design choices were made. For example, these are the first thoughts from the original lead developer of Rust, a language that was an inspiration for Swift. This is an example of the type of articles coming from professionals that want to use a state-of-the-art programming environment in a practical way. For example, Evan Miller explains how he found it more practical than Haskell, while inheriting some of its benefit in his piece, Swift impressions. Haskell is a language acclaimed for its properties, but it is often said to be used more in academic environments than by programmers for typical mobile or web software.

With Swift, what Apple wants is to do is attract and keep the best mobile developers working on iOS.

More to come on post-WWDC analysis

I will write more later on implications and expected improvements for ProcessOne and Boxcar software. I need time to think more about all the pieces of information I gathered during five days of talks and discussion with developers.

Ignite Realtime Blog: (a)Smack 4.0.0 released

8 June, 2014 - 12:03

5 months after the relase of Smack 3.4.1 the Ignite Realtime developer community is proud to annouce the first release of Smack 4, which marks a milestone in the development history of Smack. Smack has undergone a major overhaul and refactoring, including moving from Ant to Gradle and from SVN to git.


Smack 4 also includes security related fixes. Users are encouraged to update  as soon as possible.


Many people have helped to develop this release. We especially would like to thank


- Ryan Sleevi of the Google Chrome Security Team for reporting a security flaw in ServerTrustManager (SMACK-410)

- Thijs Alkemad for reporting a security flaw regarding IQ spoofing (SMACK-533, SMACk-538)

- Lars Noschinski for fixing the IQ spoofing flaws and adding support for roster versioning (SMACK-399)

- Jens Offenbach for helping making Smack an OSGi bundle (SMACK-343)


Since the API has changed in Smack 4, make sure to read the "Smack 4.0 Readme and Upgrade Guide".  A full changelog can be found in JIRA.

Thijs Alkema: CVE-2014-1361: SecureTransport buffer overflow

2 June, 2014 - 18:10

Today, Apple released a fix to CVE-2014-1361 in SecureTransport. The essence of this bug is this: the TLS record parser would interpret a DTLS record even when using normal TLS, causing a buffer overflow when parsing a record header. I reported this issue to Apple on May 28th.

To summarize, the impact of this bug is small: it can disclose 2 specific bytes of plain text to an attacker. Doing this will also cause the connection to be closed. It can also give an attacker the ability of carrying out a replay attack, with a probability of success of 2-16 (~0.0015%).


DTLS and TLS send their payloads in separate records of up to 214 bytes, where each record has a header. For TLS this header is 5 bytes: 1 byte payload type, 2 bytes TLS version number and 2 bytes indicating length of the rest of the record.

(Aside: Why every record includes two extra bytes to include the version is not exactly clear to me. I haven’t ever seen it legitimately change except during the handshake, where the client would initiate with a TLS 1.0 record, but include that it supports up to TLS 1.2, and then switch to TLS 1.2 after the server replies using that version.)

DTLS records are similar, but these are 13 bytes instead: in between the version number and the length it includes a sequence counter. Contrary to TLS, DTLS was designed to use datagrams (like UDP), so it doesn’t require reliable or in-order delivery. To still be able to decrypt records and know their intended order, the sequence counter is included on every record. TLS also uses a sequence counter (to prevent attackers from reordering messages), but it is implicit. Both parties simply count how many messages they have received or sent.

Record parsing in SecureTransport

This is how Apple’s code used to parse these records:


static int SSLRecordReadInternal(SSLRecordContextRef ref, SSLRecord *rec) { int err; size_t len, contentLen; uint8_t *charPtr; SSLBuffer readData, cipherFragment; size_t head=5; int skipit=0; struct SSLRecordInternalContext *ctx = ref; if(ctx–>isDTLS) head+=8; if (!ctx–> || ctx–>partialReadBuffer.length < head) { if (ctx–> if ((err = SSLFreeBuffer(&ctx–>partialReadBuffer)) != 0) { return err; } if ((err = SSLAllocBuffer(&ctx–>partialReadBuffer, DEFAULT_BUFFER_SIZE)) != 0) { return err; } } if (ctx–>negProtocolVersion == SSL_Version_Undetermined) { if (ctx–>amountRead < 1) { readData.length = 1 – ctx–>amountRead; = ctx–> + ctx–>amountRead; len = readData.length; err = sslIoRead(readData, &len, ctx); if(err != 0) { if (err == errSSLRecordWouldBlock) { ctx–>amountRead += len; return err; } else { /* abort */ err = errSSLRecordClosedAbort; #if 0 // TODO: revisit this in the transport layer if((ctx->protocolSide == kSSLClientSide) && (ctx->amountRead == 0) && (len == 0)) { /* * Detect "server refused to even try to negotiate" * error, when the server drops the connection before * sending a single byte. */ switch(ctx->state) { case SSL_HdskStateServerHello: sslHdskStateDebug("Server dropped initial connection\n"); err = errSSLConnectionRefused; break; default: break; } } #endif return err; } } ctx–>amountRead += len; } } if (ctx–>amountRead < head) { readData.length = head – ctx–>amountRead; = ctx–> + ctx–>amountRead; len = readData.length; err = sslIoRead(readData, &len, ctx); if(err != 0) { switch(err) { case errSSLRecordWouldBlock: ctx–>amountRead += len; break; #if SSL_ALLOW_UNNOTICED_DISCONNECT case errSSLClosedGraceful: /* legal if we're on record boundary and we've gotten past * the handshake */ if((ctx–>amountRead == 0) && /* nothing pending */ (len == 0) && /* nothing new */ (ctx–>state == SSL_HdskStateClientReady)) { /* handshake done */ /* * This means that the server has disconnected without * sending a closure alert notice. This is technically * illegal per the SSL3 spec, but about half of the * servers out there do it, so we report it as a separate * error which most clients – including (currently) * URLAccess – ignore by treating it the same as * a errSSLClosedGraceful error. Paranoid * clients can detect it and handle it however they * want to. */ SSLChangeHdskState(ctx, SSL_HdskStateNoNotifyClose); err = errSSLClosedNoNotify; break; } else { /* illegal disconnect */ err = errSSLClosedAbort; /* and drop thru to default: fatal alert */ } #endif /* SSL_ALLOW_UNNOTICED_DISCONNECT */ default: break; } return err; } ctx–>amountRead += len; } check(ctx–>amountRead >= head); charPtr = ctx–>; rec–>contentType = *charPtr++; if (rec–>contentType < SSL_RecordTypeV3_Smallest || rec–>contentType > SSL_RecordTypeV3_Largest) return errSSLRecordProtocol; rec–>protocolVersion = (SSLProtocolVersion)SSLDecodeInt(charPtr, 2); charPtr += 2; if(rec–>protocolVersion == DTLS_Version_1_0) { sslUint64 seqNum; SSLDecodeUInt64(charPtr, 8, &seqNum); charPtr += 8; sslLogRecordIo("Read DTLS Record %016llx (seq is: %016llx)", seqNum, ctx–>readCipher.sequenceNum); /* if the epoch of the record is different of current read cipher, just drop it */ if((seqNum>>48)!=(ctx–>readCipher.sequenceNum>>48)) { skipit=1; } else { ctx–>readCipher.sequenceNum=seqNum; } } contentLen = SSLDecodeInt(charPtr, 2); charPtr += 2; if (contentLen > (16384 + 2048)) /* Maximum legal length of an * SSLCipherText payload */ { return errSSLRecordRecordOverflow; } if (ctx–>partialReadBuffer.length < head + contentLen) { if ((err = SSLReallocBuffer(&ctx–>partialReadBuffer, head + contentLen)) != 0) { return err; } } if (ctx–>amountRead < head + contentLen) { readData.length = head + contentLen – ctx–>amountRead; = ctx–> + ctx–>amountRead; len = readData.length; err = sslIoRead(readData, &len, ctx); if(err != 0) { if (err == errSSLRecordWouldBlock) ctx–>amountRead += len; return err; } ctx–>amountRead += len; } check(ctx–>amountRead >= head + contentLen); = ctx–> + head; cipherFragment.length = contentLen; ctx–>amountRead = 0; /* We've used all the data in the cache */ /* We dont decrypt if we were told to skip this record */ if(skipit) { return errSSLRecordUnexpectedRecord; } /* * Decrypt the payload & check the MAC, modifying the length of the * buffer to indicate the amount of plaintext data after adjusting * for the block size and removing the MAC */ check(ctx–>sslTslCalls != NULL); if ((err = ctx–>sslTslCalls–>decryptRecord(rec–>contentType, &cipherFragment, ctx)) != 0) return err; /* * We appear to have sucessfully received a record; increment the * sequence number */ IncrementUInt64(&ctx–>readCipher.sequenceNum); /* Allocate a buffer to return the plaintext in and return it */ if ((err = SSLAllocBuffer(&rec–>contents, cipherFragment.length)) != 0) { return err; } memcpy(rec–>,, cipherFragment.length); return 0; }

head determines how many bytes the header should contain. charPtr points to the current position in the record. rec is a structure describing the record we’re parsing. ctx is the session context.

Line 195 correctly uses ctx->isDTLS, but line 309 uses rec->protocolVersion, which got parsed on line 306. This is data that just came from the network and has not been validated in any way. There are no checks to make sure rec->protocolVersion == DTLS_Version_1_0 is only true when ctx->isDTLS.

This means that an attacker can change the version number on a single record from a TLS version to DTLS 1.0 to make a user execute the if block on line 309, even though they are using a TLS connection. That might make it possible to modify the sequence counter.

Reordering attacks

The sequence counter in TLS is used to make it impossible for an attacker to remove messages, reorder messages or replay previous messages. The sequence counter is included in the MAC, which means the message will not validate when it isn’t in its original place in the sequence. Due to the bug in the code above, the attacker may be able to modify this sequence counter. What an attacker can do with that is hard to determine: it depends a lot on the exact fragmentation of the payload into records.

In HTTPS, for example, an attacker may try to make some JavaScript execute differently, but if the entire script fits in one record then there’s not much an attacker could do. The most efficient way to send webpages or scripts would be to make as few records as possible, as padding and MAC add overhead per record. This means fragmenting the data every 214 bytes = 16 KiB (except for a bit of room for the MAC). By comparison, the current version of jquery is 82 KiB. That would fit in 6 records, giving any attacker very few options to shuffle those fragments around, many of these will probably not even parse as valid JavaScript.

In more real-time protocols like IRC or XMPP (yes, of course I have to bring up XMPP again), the fragmentation is a lot easier to understand: these will include a few complete protocol packets within each record (often just 1). Having a malicious impact here will be a lot easier: an attacker would be able to drop a single chat message, retransmit one, reorder them, etc.

Rewriting the sequence number

Trying to exploit this, I quickly ran into the following problem: only 5 bytes of the record had been copied from the socket, so the SSLDecodeUInt64 call will read 2 bytes from the record, but 6 bytes past that too. This does makes it possible to make sure the epoch matches (the two highest bytes of the sequence number), but the 6 next bytes are “random” data.

Looking a little closer, the next 6 bytes didn’t turn out to be random at all. The buffer records are read into gets reused (except when a record has too much payload to fit in the current buffer, then a new one is allocated) and decryption of the record happens in-place in this buffer. So when I tried to exploit this using a HTTPS server which had previously sent a reply starting with HTTP/1.1 200 OK, Safari ended up interpreting HTTP/1 as the sequence number. The length field of the record should follow the sequence number, so it interpreted .1 as its length.

TLS 1.0

I tried a lot of variations, setting up some plaintext in the buffer first and then trying to reinterpret that as the sequence counter, until finally I realized what I was trying to do wasn’t possible with TLS 1.0: all the ciphers I was trying used more inter-record state than just the sequence counter. CBC mode means the decryption of every record depends on the ciphertext of the previous record, so reordering would never work. RC4 keystreams are also inherently statefull. As TLS uses MAC-then-Encrypt (MtE), these records will decrypt to gibberish and then fail the MAC. If TLS had used Encrypt-then-MAC (EtM) here (which a lot of cryptographers nowadays consider the better choice), the MAC would have succeeded, after which the record would have decrypted to gibberish. That gibberish would’ve been passed to the application, as the TLS layer would not have been able to detect anything wrong with it.

TLS 1.1+

TLS 1.1 and TLS 1.2 don’t have that problem: these add an explicit IV to every record to prevent attacks like BEAST. For compatibility with TLS 1.0, this is usually implemented by prepending a block of random data to the plaintext and including that in the encryption. The IV that is used to encrypt this new first block doesn’t matter: it only influences the plaintext of the first block, which is deleted by the receiver after decryption. It doesn’t even need to be the case that the receiver decrypts the first block to the same thing as the sender used. So here every record can be decrypted independently, even when inserted at a random other position in the sequence. In practice, the IV that is used as the IV for the first block is often still the ciphertext of the last block of the previous record, as that makes it easier to be compatible with TLS 1.0 while not being vulnerable to BEAST.

However, this also meant that the sequence number was no longer the ASCII encoding of HTTP/1 (or the first 6 bytes of whatever record was last), but it is now the decryption of the IV block. As this block gets chosen randomly and the server and client don’t even need to decrypt it to the same thing, trying to influence this block to contain just the sequence number I want turned out to be impossible.

My next thought would be to send a record with a wrong epoch first, which would be used to fill the buffer with the data I need and then send another record with a DTLS header that would be used to overwrite the sequence counter. In DTLS, the epoch is indicated by the two upper bytes of the sequence counter. Records with an epoch different from the epoch of the current sequence counter are skipped (decryption or authentication isn’t attempted).

However, this just moved the problem backwards: the length of this new record is still taken from the data still in the buffer, so the decrypted IV of the previous record. Even though this new record will not be decrypted, SecureTransport must read it completely first, and I don’t know what length it expects. Guessing would have a 1 in 216 chance of succeeding, which is large cryptographically speaking, but not quite practical. It might be possible to increase this chance by repeating the inserted record over and over, but then the attacker can only insert one record, as the next copy will fail to decrypt.


I believe AES-GCM would be vulnerable to this, as it uses the sequence number as an implicit IV, though I haven’t checked. While SecureTransport has an (at least partial) implementation of AES-GCM, it wasn’t advertised by Safari, so I’m assuming its unfinished.

Leaking bytes

Another avenue of exploitation would be to try to retrieve some information about the plaintext still in the buffer. As I mentioned in my HTTPS example .1 from HTTP/1.1 200 OK was interpreted as the length of the next record. The ASCII representation of .1 interpreted as a number gives 11825. This means SecureTransport will try to read 11825 more bytes before starting to decrypt it (which will then fail the MAC, causing it to send an alert and close the connection). We can also do this the other way around: we write bytes one by one until SecureTransport closes the connection and from that we will know the 7th and 8th byte of the plaintext of the previous record!

However, the value of the two bytes has to be less than the maximum record size of 214 (while it can be up to 216), as otherwise SecureTransport will reject it for being too large. This means that the first character must have an ASCII representation of less than @, which means it can’t be any of the upper- or lowercase letters, but numbers and a few other punctuation characters would work.

Closing thoughts

After heartbleed, this is another bug that exploits a DTLS code path that should never be used when using TLS. Impact is even similar too: disclosing some contents of the other side’s memory. However, this is only limited to 2 bytes, while heartbleed could retrieve 64 KiB per heartbeat. I guess DTLS has its uses, but maybe implementors should consider whether covering both DTLS and TLS in one library is worth the extra complexity of security-critical code.

A discovery that surprised me is the way SecureTransport deals with its internal buffers. The buffer records are read into and where the result of their decryption is stored are never erased, there’s only malloc and free. Buffers grow when they need to receive a larger record, but they never decrease in size again for as long as the connection is open. This means long-lived TLS connections waste a lot of memory when they receive a single large record. The plaintext of that record will stay in memory for as long as the connection is open.