We are experiencing a high growth in the number of web applications being developed. This is happening mainly because the web is going into a new phase, called programmable web, where several webbased systems make their APIs publicly available. In order to deal with the complexity of this emerging web, we de ne a notion of social machine and envisage a language that can describe networks of such. To start with, social machines are de ned as tuples of input, output, processes, constraints, states, requests and responses; apart from de ning the machines themselves, the language de nes a set of connectors and conditionals that can be used to describe the interactions between any number of machines in a multitude of ways, as a means to represent real machines interacting in the real web. This work presents a preliminary version of the Social Machine Architecture Description Language (SMADL).
Software systems are built upon programming languages. A programming
language is a notation for expressing computations (algorithms) in both machine
and human readable form. Appropriate languages and tools may drastically
reduce the cost of building new applications as well as maintaining existing ones
[
In the context of programming languages, a Domain-Speci c Language (DSL)
is a language that provides constructs and notations tailored toward a particular
application domain [
However, in software engineering several di erent artifacts are developed
besides code and one of the most important is the software architecture. Most
developers agree that architecture is needed in some way, shape, or form, but, they
can't agree on a de nition, don't know how to manage it e ciently in nontrivial
projects, and usually can't express a system's architectural abstractions precisely
and concisely [
These answers mention aspects that are actually related to a system's architecture, but none of them represent an unambiguous and/or \formal" description of a system's core abstractions. Indeed, it is not surprising because, although there are languages that directly express software architectures, they are not quite common among software developers.
In order to better de ne software architectures, it is worthy using DSLs and taking advantage of their expressiveness in a limited domain. Our proposal relies on top of an Architecture Description Language (ADL) for describing web based software systems in terms of Social Machines, a new concept developed by our research group which tries to increase the abstraction level for comprehending the web. Next, we present the context and the details about Social Machines. 2
The traditional concept of software has been changing during the last decades.
Since the rst de nition of a computing machine described by Turing in [
Recently, we all can clearly see that a new phase is emerging, the web \3.0", the web as a programming platform, the network as an infrastructure for innovation, on top of which all and sundry can start developing, deploying and providing information services using the computing, communication and control infrastructures in a way fairly similar to utilities such as electricity.
An overview of this Web 3.0 scenario can be seen in the ProgrammableWeb
website1. It gathers around 6500 publicly available APIs and more than 6700
mashups using them (last visit in July 2012). Although there have been many
studies about the future of the internet and concepts such as web 3.0,
programmable web [
Furthermore, the Web concepts, as we know, are recent enough to represent
many serious di culties while understanding their basic elements and how they
can be e ciently combined to develop real, practical systems in either personal,
social or enterprise contexts. Therefore, we developed a new concept called Social
Machine (SM), in order to provide a common and coherent conceptual basis
for understanding this still immature, upcoming and possibly highly innovative
phase of software development. SM concept was rstly conceived in [
So, we de ne a SM as a tuple, as following:
SM = <Rel, WI, Req, Resp, S, Const, I, P, O>
In general, a SM represents a connectable and programmable entity
containing an internal processing unit (P) and a wrapper interface (WI) that waits for
requests (Req) from and replies [with responses (Resp)] to other social
machines. Its processing unit receives inputs (I), produces outputs (O) and has
states (S); and its connections de ne intermittent or permanent relationships
(Rel) with other SMs. These relationships are connections established under
speci c sets of constraints (Const). Our goal with this concept of a Social
Machine is not to formally describe software services as can be seen in [
The idea behind Social Machines is to take advantage of the networked environment they are in to make it easier to combine and reuse exiting services from di erent SMs and use them to implement new ones. Hence, we can highlight some of its main characteristics, as following: Sociability, Compositionality, Platform and Implementation independency, Self-awareness, Discoverability and last, but not least, Programmability.
There may be di erent types of social machines, but one way to classify them is through the simple taxonomy shown in Figure 2, based on the types of interactions they have with each other, as follows: { Isolated - Social Machines that have no interaction with other Social Machines; { Provider - Social Machines that provide services for other Social Machines to consume; { Consumer - Social Machines that consume services that other Social Machines provide; { Prosumer - Social Machines that both provide and consume services.
In this work, we envisage an Architecture Description Language that can describe networks of SMs. Apart from de ning the machines themselves, the ADL de nes a set of connectors and conditionals that can be used to describe the interactions between any number of machines in a multitude of ways, as a means to represent real machines interacting in the real web. Details are presented next. 3
This work is an attempt to answer the following research questions: \Is it possible to integrate diverse web applications using a standard architecture description language? ". In order to answer it, this work purposes a new ADL for de ning social machines: SMADL.
Social Machines can be connected (or establish a relationship) in basically two phases: in the rst phase, the SMs must nd each other, and a there must be a SM registry service much likely Internet DNS; in the second phase the SMs actually connect to each other and exchange information for a limited period of time. The SM registry service is out of the scope of this proposal. We are assuming SMs can nd each other without much e ort.
In order to comprise these two phases, SMADL is composed by two minilanguages: { VCL (Visitor Card Language): presents the externally visible properties of a SM, i.e., which requests and responses it accepts, which types of inputs/outputs it handles, if an internal state is maintained and/or how many requests it can handle per amount of time. A vCard is provided by the SM registry to the consumer SM, so it can decide if the relationship is interesting or not. Business issues are also present in a SM vCard, such as, billing information and service level agreements. Nowadays, popular web APIs do not make available such business information in a programmatic way. Usually, there are only few lines in reduced font size contracts mentioning that important information. { WIL (Wrapper Interface Language): we are assuming every SM has a vCard with which the wrapper interface is fully complaint. This language is responsible for actually connecting SMs, establishing pre and post conditions, applying di erent connectors in a SM composition, and implementing business rules associated with a given set of SMs. Our proposal is to use WIL not as substitute for the currently available technologies. Instead it increases the level of abstraction of these technologies, freeing the programmer to concentrate on business issues of the SM relationships.
To understand better how these mini-languages are used, Figure 3 shows the steps for establishing a relationship between two SMs, as following: 1. Initially, the requester SM, in this case represented by Evernote (upper left icon), searches for some SM registered as a micro blog, in our case Twitter (upper right icon). Note that Evernote presents its vCard to DNS while searching for a service, once the responder may or may not accept connections with that speci c SM. 2. The SM DNS nds Twitter and requests its vCard. 3. Twitter responds with its vCard, accepting the relationship. 4. The SM DNS replies back to Evernote with the Twitter vCard, which includes its address. 5. Using Twitter vCard and knowing its wrapper interface, Evernote establishes a relationship with Twitter, following all conditions imposed.
There are several popular technologies for integrating web based or service
oriented systems, such, REST [
SMADL is being developed on Xtext language workbench [
In our study we could enumerate 30 publications directly related to ADL.
Amongst them, only two of them mention the Web domain, both from 2010. In
the rst one [
In addition, practical examples, such as Yahoo! Pipes2 and IfThisThenThat 3 can be seen as related work. The former uses a graphical tool for customizing data ows from di erent sources. The latter allows end users to program the web based on pre-de ned events red by a set of channels, for example, if someone tags you on a given social network (channel 1), then save this photo in the person's virtual drive (channel 2). The user can choose among di erent events from di erent channels, which are, in practice, websites that make available their APIs. Our work is an attempt to be a completely di erent way to program
the Web, not based on pre-de ned parameters. The idea behind SMADL is to actually de ne every public API in the Web and the relationships among them. This way, composition possibilities for several SMs can be in nite.
As can be seen, this is a relatively new research eld and we believe we can make a considerable contribution by establishing the concept of a Social Machine and developing an ADL for supporting it. 5
This work presents SMADL \The Social Machines Architecture Description Language" as a possible solution for modeling web-based software systems.
In general, a Social Machine (SM) represents a connectable and programmable entity containing an internal processing unit (P) and a wrapper interface (WI) that waits for requests (Req) from and replies [with responses (Resp)] to other social machines. Its processing unit receives inputs (I), produces outputs (O) and has states (S); and its connections de ne intermittent or permanent relationships (Rel) with other SMs. These relationships are connections established under speci c sets of constraints (Const).
Our main goal is to use SMADL to describe SM relationships and then we can have one unique way to program the web, independently of what technology/platform is being used. The current version (alpha) of SMADL generates code for REST based apps, as it is becoming the most popular on the web, adopted by big players such as Facebook and Google. Nowadays, we are working on several sample apps which have their architecture written in SMADL. These apps are basically consumer SMs, it means they do not make available public features, but only consumes other public APIs. These apps are going to be distributed as open source.
Our next steps include writing fully prosumer social machines, i.e.
applications that connects to others, process their data, and someway make this data
available for others to consume. At this phase, we are planning to perform an
experiment, following the methodology described in [
Acknowledgments. This work was partially supported by the National Institute of Science and Technology for Software Engineering (INES ), funded by CNPq and FACEPE, grants 573964/2008-4, APQ-1037-1.03/08 and APQ1044-1.03/10 and Brazilian Agency (CNPq processes number 475743/2007-5 and 140060/2008-1).