The pace of technology change has accelerated in the past decade. Conceptually similar technologies are introduced on nearly a daily basis. On one hand, IT experts call for applying the most modern approaches and technologies to software projects, on the other, companies su er from liabilities to a technology used in their legacy solutions. This seems to result in a disturbing situation when a speci c technology of an ongoing software project becomes legacy almost before the project successfully hits a production. This poses a continual challenge for software development, and e ective ways of technology transition are sought. A ordance-Driven Assembling (ADA) represents such an e ort from the standpoint of enterprise engineering theories. In this paper, we formulate a high-level architecture of a software system based on ADA. We demonstrate the architecture on an example of an object-oriented system. We evaluate the qualities of such architecture from the perspective of evolvability using Normalized Systems Theory, and we formulate conclusions on potential of this approach.
In the past decades, a range of new software technologies and frameworks have been developed to provide various solutions to software development challenges. For instance, we can nd tens of frameworks and libraries for Dependency Injection, Object-Relational-Mapping, Service-Bus, we can dabble with hundreds of User Interface (UI) technologies, or rummage thousands of controls in JavaScript. Although there are considerable di erences among them, many of them address the same class of problems. At the same time, it feels like the pace at which these frameworks are introduced is accelerating.
We commonly refer to the famous Moore's Law which observes this
phenomenon in computing. This widely known exponential doubling of transistors is
powering the advances in the past ve decades. However, it seems there is a
similar common force driving information technology forward. In [
If Kurwzeil is right, we can expect that more and more frameworks and
libraries will be introduced even quicker. This brings a serious problem to deal
with legacy software, i.e., software which has been developed using already
outdated technology. Due to the technology pace, the software systems can thus
become legacy almost before hitting a production { as the pace of
development technology increases, so too does the pace of technology obsolescence [
In this paper, we focus on technologies used for UI development where the
problem of a technology transition is arguably the most observable in the
proliferation and dynamics of UI frameworks. For instance, the JavaScript community
coined a term the \JavaScript fatigue". Regardless the progress and popularity
of JavaScript, developers report that they are dealing with a fatigue which refers
to an inability to keep up with the latest libraries, there is a fear of becoming
obsolete and outdated due to a constant change of the ecosystem and an
overwhelming number of choices [
We argue that by following certain architectural patterns in a software
system, we can adapt software artefacts to the latest technologies more e ciently.
In this paper, we elaborate on the ADA (A ordance-Driven Assembling)
approach that is founded in formal EE (enterprise engineering) theories. In [
To measure an improvement in helping technology transition, we work with
the term \evolvability". We understand \evolvability" in terms of Normalized
Systems Theory (NST) [
When migrating a software from one technology to another, we su er from
incompatibilities between them. In practice, we usually tackle the problem of
moving to another framework, or library based on the same platform or language,
e.g., .NET, Java, C++, JavaScript. Thus, we typically identify parts which can
be left untouched while the rest is adopted to the new technology. This rises a
question of what are the concepts that can be easily reused, and which make it
harder to move. Additionally, in which way we have to capture these concepts
in a software system in order to optimize it for future upgrades. In [
In [
When building software systems based on this de nition, we must be able
to identify so-called ADA-users, describe their ADA-purposes, and based on
that build the nal solution composed of ADA-components. All of these ADA
artifacts are formally de ned in [
The COPS company has implemented an application framework corima. It is an example of ADA in practice. The system corima is capable of hosting applications in various business domains1. By now, the biggest application suite is focused on banking and corporate treasury management. It encompasses around 50 1 Under the business domain, we understand the area of a business, e.g., nance, health-care, etc. data-centric web and desktop applications used across tens of customers mainly in Europe.
The system corima implements ADA concepts captured in Fig. 1. We evaluated its code-base in order to measure the impact of a technology or domain transition.
Our preliminary calculations show that if the change-driver is a technology (e.g., a move to another UI framework), the impact of such a change is bounded. We only need to change a code which corresponds to the Mapper and Component Repository in Fig. 1. This only makes 2.2% of the entire code-base.
On the other hand, if the change driver is in the domain. The adjustment of Purpose and User repository is bounded to 1.7% of the whole code-base and it won't change with the amount of new applications. 5
In this paper, we started by observing that the pace of introducing new technologies is exponentially accelerating, posing a serious challenge for the software industry with systems becoming quickly legacy ones much sooner. We addressed this challenge by showing how ADA can ease technology transitions.
We formulated a possible high-level architecture for software systems following ADA and demonstrated it on a case study of a real system. Using NST, we evaluated that ADA improves the evolvability of software in situations of technological transitions by putting clear boundaries of change impact.
Although our measurements indicated the potential of ADA, further statistical evaluation on a large number of cases is as subject of a future work.