Nowadays, mobile devices constitute the most common computing device. This new computing model has brought intense competition among hardware and software providers who are continuously introducing increasingly powerful mobile devices and innovative OSs into the market. In consequence, cross-platform and multi-device development has become a priority for software companies that want to reach the widest possible audience. However, developing an application for several platforms implies high costs and technical complexity. Currently, there are several frameworks that allow cross-platform application development. However, these approaches still require manual programming. My research proposes to face the challenge of the mobile revolution by exploiting abstraction, modeling and code generation, in the spirit of the modern paradigm of Model Driven Engineering.
Context Nowadays, mobile devices constitute the most common computing device. A vast array of features has been incorporated into those devices to address the di erent demands of users spanning from games to serious business. Today mobile devices are as powerful as desktop computers in terms of their computing capabilities. This new computing model have brought intense competition and innovation among devices, OSs, and application providers.
Problem Even though the mobile OS market is beginning to mature and
consolidate, most researches concur that it is unlikely that a single vendor will
dominate the future mobile-centric world [
Moreover, the vastness and diversity of mobile devices and operating systems available on the market oblige companies to produce and deploy the same app several times, once for each of the di erent mobile platforms. Unfortunately, cross-platform and multi-device development is a barrier for today's IT solution providers, especially SMEs, due to the high cost and technical complexity of targeting development to a wide spectrum of devices, which di er in format, interaction paradigm, and software architecture.
Currently, there are several frameworks implementing di erent
methodologies for cross platform application development (Web, Hybrid, Interpreted and
Cross Compiled): examples include PhoneGap(Cordova) 1, Appcelerator
Titanium2, and Xamarin3. Unfortunately, these approaches still require manual
programming which yields to high risks of errors, inconsistencies and ine ciencies.
Relevance The number of apps that are available in the online markets has
reached unseen numbers. In fact, by July 2014, the Google Play store counted
1.3 million of available apps while Apple's App Store counted 1.2 million. In
parallel with these numbers, the market also expects an increase in the number
of global smart-phones users, which is expected to surpass 2 billion by 2016 [
The paper is organized as follows: Section 2 describes the problem that my research intends to solve; Section 3 reviews the related work; 4 presents the proposed solution; Section 4 presents the methodology and tools that will help to use the proposed solution; and Section 6 presents the preliminary work, the future works and the contributions expected from my research.
Cross-platform and multi-device development is a barrier for today's IT solution providers, especially SMEs, due to the high cost and technical complexity of targeting development to a wide spectrum of devices, which di er in format, interaction paradigm, and software architecture. The challenges of mobile apps developments that my research attempt to address are described in detail in the next paragraphs.
P1: Platform. The market of mobile operating systems is fragmented and
rapidly changing. The diversity of OSs available on the market oblige software
developers that want to reach a large audience of users to develop their apps for
each platform (at least for the most competitive ones such as Android, iOS, and
Windows Phone [
P2: Di erent Front End Requirements. From the user interactions
perspective, mobile apps are expected to support a wider set of interactions that
are captured by means of a tactile surface (interaction through a set of gestures
like taps and swipes ) and through the di erent sensors that are packed into
the device (sensor-based interactions like rotate and shake ). Moreover, front-end
design of mobile apps must consider the size constraints imposed by the
characteristics of the screens of modern mobile devices. In addition, the mobile apps
must adapt to changes of the context (the communication network, the battery
level of the device and the environment surrounding the user) to deliver the most
e cient interface [
P3: Resource Scarcity. Even though the mobile OSs provide the optimization strategies to cope with the scarcity of resources on mobile devices (like memory and storage, battery, and the instability and diversity of the communication networks), mobile apps need to be able to receive system noti cations (such as the battery level and new networks availability) and react appropriately to them in order to provide a consistent and reliable user experience.
P4: Device Diversity. Currently, the market o ers several families of mobile devices such as tablets, smart-phones and emerging smart-watches. Each of them can show di erent amounts of information, uses particular navigation patterns and has a diverse set of sensors available at run-time. These di erences imply that software developers have to create applications that can either adapt to the speci c device in which they are running, or create di erent versions of the application, each of them targeting a speci c device family. 3
This research is the rst one that attempts to compare various model-driven strategies for cross-platform and multi-device mobile apps development with aim of providing guidelines to the developers who need to adopt MDD approach in their mobile apps development process. Thus, this section assesses the existing works that apply the MDD approach to the development of mobile apps in a broad sense.
Those works can be divided into two di erent clusters. On one hand we
encounter a corpus of research that apply model-driven techniques to specify
application interfaces and user interaction (in a broad sense) for multi-device
UI modeling. Among them we can cite: TERESA(Transformation Environment
for inteRactivE Systems representations) [
My research will leverage on the IFML language which has been recently
adopted as a OMG standard. In particular my research provided a mobile
extension [
On the other hand we nd a collection of works that proposes MDD solutions for the development of cross-platform mobile apps. This cluster of researches can be further divided into three groups depending on whether they produce native, hybrid or web-based apps.
In the rst group we encounter projects like MD2 [
The second group concerns solutions that generate hybrid apps to address the cross-platform issue. As part of this group we can mention Applause4, a domainspeci c language and a set of code-generators to produce mobile apps for iPhone, Android, Windows Phone on top of Google App Engine. As mentioned before, my solution will consider the generation of both native and hybrid apps and will provide a graphical syntax, which are both missing in Applause.
Finally, the third group contains MDD proposals that generate web-based mobile apps. A good representative of this category is Mobl5, an open source language designed to speed up building mobile apps. Mobl o ers a concise language to build native-feeling web apps for mobile system. Even though this approach towards cross-platform development is not considered in my solution, it could be seen as a complementary strategy suitable for speci c scenarios where the native or the hybrid solutions do not bring any additional value. 4
I propose to face the challenge of cross-platform multi-device mobile apps development by applying a model driven development (MDD) approach. I will use the model driven architecture (MDA)6 as a reference framework to illustrate the proposed solution. MDA de nes models at three di erent levels of abstraction: Computation Independent Models (CIM), Platform Independent Models (PIM), and Platform Speci c Models (PSM). A set of mappings between each level and the subsequent one can be de ned through model transformations. Every CIM can map to di erent PIMs, which in turn can map to di erent PSMs but many other combinations can be followed, for instance skipping one of the levels.
My research proposes to address the problem of cross-platform mobile apps development through four code generation alternatives depicted in Figure 1. The code generators input the models that describe the app requirements and output the app code. The domain model and the business logic of the app are de ned through UML while the app front-end is de ned through IFML. In contrast I will deal with the issue of multi-device mobile apps development through model transformations that would be applied at the PIM level. Section 4.1 is dedicated to the issue of cross-platform mobile apps development (P1, P2 and P3 ) while the solution to the problem of multi-device development (P4 ) will be presented separately in Section 4.2. 4.1
When following the MDD approach, several code generation strategies are possible depending, both, on the abstraction level to be used when modeling the application and the abstraction level of the code to be generated.
In this research, I will analyze all four alternatives for cross-platform mobile apps development ( gure 1) with the aim of providing apps developers the guidelines to choose the right MDD approach for them according to their particular requirements. To reach this goal, the pros and cons of each of those options will be studied.
Platform Independent Model (PIM)
M2T Native Code (1)
M2M
PSM
M2T Native Code (2)
M2T Cross Platform
Code (3)
M2M Cross Platform
Framework Specific Model (FSM)
M2T Cross Platform
Code
(4)
(1) PIM-to-Native Code. By following this option, the app requirements shall
be speci ed through a Platform Independent Modeling Language such as
mobile IFML [
This option simpli es the creation of a cross-platform app (P1) because it generates the code for a cross-platform solution. For example to create an app that will run both on Android and iOS with PhoneGap as the crossplatform framework a software designer will start by creating the Platform Independent Model of the application. Then, using a single M2T transformation he will produce the HTML5, CSS and Javascript code required by PhoneGap. Finally, using the build tools o ered by the framework the code will be transformed into the binary les required by each of the target platforms, which in our case are the ipa le for iOS and apk le for Android. A Platform Independent Modeling language shall allow the modeling of the speci c Front End requirements that characterize mobile applications (P2), as well as the interactions that occur between the app and the OS (P3). (4) PIM-to-Framework Speci c Model (FSM)-to-CPC. With respect to the previous option, this approach introduces the FSM which gathers the information regarding the cross platform framework (such as PhoneGap, AppCelerator Titanium, and Xamarin) used to produce the apps. FSM is a PSM in which the Platform in the MDA terminology, is actually a crossplatform framework for mobile apps development. In this case the PIM is rst transformed into the FSM which re nes the initial model adding the cross-platform framework speci c details that are not captured at the PIM level. Once the FSM has been produced, a simple code generator transform that model into the code required by the cross-platform framework. The introduction of FSM level requires to provide a FSM modeling language. Similarly to the previous option, the problem of front-end requirements (P2) and resource scarcity (P3) are addressed at modeling level while the problem of platforms diversity is achieved by the cross-platform framework itself.
All the options mentioned so far require a modeling language allowing the speci cation of application requirements in a platform independent manner. In this research I will use an OMG standard, the Interaction Flow Modeling Language (IFML) as a reference PIM language. In particular this research will de ne a mobile extension of IFML to allow the modeling of the speci c Front End requirements that characterize mobile applications, as well as the interactions that occur between the application and the OS. The PSM languages required by options (2) and (4) will be de ned as Platform Speci c Extensions of the mobile extension of IFML mentioned before. To clarify the solution, Figure 2 shows a diagram with the proposed modeling stack. 4.2
The solution I propose to address the issue of device diversity (P4) is based on the following assumption: the characteristics of di erent devices of the same family (tablet, smart-phone, smart-watch, etc) do not change drastically. For instance di erent smart-watches are assumed to have more or less the same screen dimensions and di erent smart-phones are assumed to have have roughly the same sensors.
To deal with this issue, I propose a strategy based on a set of model to model transformations (M2M) that are applied over the same general model of the mobile application. Each of these transformations will produce a new model that describes the shape the application will have in a particular device family (Figure 3). For example, when designing an application that should be used in a tablet, a smart-phone and a smart-watch, a software designer will rst create a single general model of the application. Then, he will use a particular M2M transformation to generate a version of the model that is suitable for phones. He will then repeat the same process using di erent M2M transformations to obtain the model for the tablet and the watch. At the end of the day, the software designer will have four di erent Platform Independent Models, that can be transformed into running code following any of the previously discussed strategies.
It is important to highlight that the aforementioned M2M transformations could be de ned using two di erent approaches. The rst approach is based on PIM for Tablet
M2M
M2M
PIM for Smart-Phone
M2M
PIM for Smart-Watch
M2M
PIM for
…
the de nition of a xed set of transformation rules that given a general model,
are able to produce the model suitable for a particular device family. In the
second approach, the M2M transformation rules are de ned by the software
designer with the support of a model editor that records at each step the
modi cations he applies over the general model to obtain the model of a particular
device family [
The results of a comparative study among di erent MDD approaches to crossplatform and multi-device mobile apps development (Section 4) will constitute guidelines for apps developers to choose a MDD approach that ts their requirements.
Front-end design of mobile apps is a complex task, the content and the navigation among them must be well designed at the purpose of exploiting at best the limited space available. Providing design patterns for both the content organization and navigation could help application designers to nd solutions to common design challenges and to reuse them. My research will provide a set of design patterns as to illustrate and simplify the modeling of mobile app using the de ned method and tools. 6
Preliminary Work. My research starts from a deep state-of-the-art analysis
on mobile apps development in a wide sense on model driven approaches in
particular. The current implementations include:
{ A platform Independent Modeling language for mobile apps [
{ A set of rst prototypes of code generators both for native platforms and cross-platform frameworks; { Initial validation through developed mobile apps that includes Instangram,
CamScanner, and Foursquare.
The future works include:
{ Model-driven analytic. I will study how model-driven techniques can be
combined with existing analytic in order to provide a rich message from web and
mobile apps monitoring;
{ Multi-Devices Development. The issue of model to model transformations
(M2M) mentioned in Section 3.2 needs a deep investigation to understand
whether is better to rely on application independent and xed M2M rules,
application dependent rules or a combination of both [
Expected Contributions. My research is expected to make three main contributions. Firstly, it will give a comprehensive overview of current MDD approaches for cross-platform app development. Secondly, it will provide a framework of criteria for evaluating MDD approaches to mobile apps development. The proposed criteria could be used for future assessments. Thirdly, it will provide decision advice allowing developers to choose a MDD approach that better ts their requirements.
In practice the contribution of my research can be appreciated from two di erent
perspectives. First, from the point of view of the software development
companies who could appreciate a MDD approach to cross-platform mobile
applications development that can reduce the technical complexity and the development
costs [