0% found this document useful (0 votes)
17 views10 pages

Low-Code Development and Model-Driven Engineering Two Sides of

This paper discusses the rise of low-code development platforms (LCDPs) and their relationship with model-driven engineering (MDE). It compares and contrasts the two approaches, highlighting their common goals and differences, while also analyzing their strengths and weaknesses. The authors propose potential areas for collaboration and knowledge exchange between the two communities.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
17 views10 pages

Low-Code Development and Model-Driven Engineering Two Sides of

This paper discusses the rise of low-code development platforms (LCDPs) and their relationship with model-driven engineering (MDE). It compares and contrasts the two approaches, highlighting their common goals and differences, while also analyzing their strengths and weaknesses. The authors propose potential areas for collaboration and knowledge exchange between the two communities.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 10

Software and Systems Modeling (2022) 21:437–446

https://doi.org/10.1007/s10270-021-00970-2

EXPERT VOICE

Low-code development and model-driven engineering: Two sides of


the same coin?
Davide Di Ruscio1 · Dimitris Kolovos2 · Juan de Lara3 · Alfonso Pierantonio1 · Massimo Tisi4 · Manuel Wimmer5

Received: 2 September 2021 / Revised: 8 December 2021 / Accepted: 9 December 2021 / Published online: 11 January 2022
© The Author(s) 2022

Abstract
The last few years have witnessed a significant growth of so-called low-code development platforms (LCDPs) both in gaining
traction on the market and attracting interest from academia. LCDPs are advertised as visual development platforms, typically
running on the cloud, reducing the need for manual coding and also targeting non-professional programmers. Since LCDPs
share many of the goals and features of model-driven engineering approaches, it is a common point of debate whether low-
code is just a new buzzword for model-driven technologies, or whether the two terms refer to genuinely distinct approaches.
To contribute to this discussion, in this expert-voice paper, we compare and contrast low-code and model-driven approaches,
identifying their differences and commonalities, analysing their strong and weak points, and proposing directions for cross-
pollination.

Keywords Low-code development · No-code development · Model-driven engineering

1 Introduction code. The main aims of LCDPs are to reduce the development
and maintenance effort required to deliver and operate cer-
Low-code development platforms (LCDPs) are on the rise, tain types of applications and to enable digital-savvy citizen
with an increasing number of cloud vendors, such as Google, developers who lack or have limited programming expe-
Microsoft, and Amazon, offering solutions for developing rience to contribute to the software development process
and operating complex software applications with little or no directly.
As model-driven engineering (MDE) [3] has similar aims,
Communicated by Bernhard Rumpe. there is an ongoing debate on how low-code software devel-
opment is different from model-driven engineering and to
B Davide Di Ruscio
[email protected] what extent work carried out in the field of MDE is directly
Dimitris Kolovos
transferable to LCDPs [6]. In this paper, we aim at clarifying
[email protected] the commonalities and differences between both approaches.
Juan de Lara
We argue that while the two approaches share similar high-
[email protected] level aspirations, there are also differences, for instance, not
Alfonso Pierantonio
all model-driven techniques aim at reducing the amount of
[email protected] code needed to implement software solutions, and not all
Massimo Tisi
low-code approaches are model-driven.
[email protected] The rest of the paper is organised as follows. In Sect. 2, we
Manuel Wimmer
summarise the history of the low-code movement we have
[email protected] seen so far as a discussion basis for the subsequent sections.
In Sect. 3, we provide an overview of typical low-code devel-
1 Università degli Studi dell’Aquila, L’Aquila, Italy opment processes and tools that LCDPs offer. In Sect. 4, we
2 University of York, York, UK contrast and compare the principles and practices of low-
3 Universidad Autónoma de Madrid, Madrid, Spain code development and MDE. In Sect. 5, we discuss possible
4 IMT Atlantique, LS2N (UMR CNRS 6004), Nantes, France
reasons behind the increasing adoption of LCDPs. In Sect. 6,
5
we identify lessons that the two communities can learn from
Johannes Kepler University Linz, Linz, Austria

123
438 D. Di Ruscio

each other. Finally, Sect. 7 summarises and concludes the


paper.

2 The history of low-code development

Since the exact meaning and industrial relevance of low-code


development is controversial, we believe it is important to
start the paper by presenting the original definitions of the
term and the main events of its history.
The past decades have been marked by several indus-
try trends aiming at reducing the amount of hand-crafted
code required to produce software such as 4GLs and CASE
tools in the 1980s [28], Rapid Application Development in
the 1990s [29], End-User Development in the 2000s [26],
and MDE in the last two decades [54]. The first use of
the term low-code can be traced back to the market anal- Fig. 1 Major events in low-code history
ysis firm Forrester in 2014 [43] (cf. Fig. 1), where low-code
development platforms (LCDPs) were defined as “platforms
that enable rapid delivery of business applications with a
minimum of hand-coding and minimal upfront investment The year 2017 noted the start of a series of acquisitions for
in setup, training, and deployment”. It is interesting to LCDP vendors [46]. Appian started an initial public offering
note that this report identified the LCDP segment as spe- in May 2017, and in 2018 its market valuation nearly reached
cific to the production of business applications, such as $2 billion. In July 2018, OutSystems received investments of
software for enterprise resource planning, customer rela- $360 million. In August 2018, Siemens announced the acqui-
tionship management, business process management, and sition of Mendix for $730 million [46]. In 2017, Forrester
other productivity-enhancing applications. In 2016, Forrester estimated a global market size for LCDPs of $3.8 billion.
detailed the successful application domains for LCDPs in Forrester also periodically surveys developers about LCDP
four specific application scenarios, i.e. database, request- usage1 : In 2018, 23% of developers reported using low-code
handling, process, and mobile-first [49]. platforms, and another 22% planned to do so within a year
The definition has evolved, and in 2017, Forrester pro- [48]. In 2019, 37% of developers were using or planning to
vided a more detailed version, characterizing LCDPs as use low-code products [32].
“products and/or cloud services for application development In 2021, most large cloud providers offer LCDPs within
that employ visual, declarative techniques instead of pro- their cloud-based solutions. Microsoft was among the first
gramming and are available to customers at low- or no-cost to embrace the trend by releasing its Power Apps LCDP
in money and training time to begin, with costs rising in pro- in November 2016. In January 2020, Google acquired the
portion of the business value of the platforms” [45]. The focus LCDP provider AppSheet and made it its flagship low-
here is on visual interfaces and declarative techniques, with code solution. In June 2020, Amazon released Honeycode, a
Forrester especially emphasizing visual WYSIWYG devel- LCDP for web, and mobile application development.
opment and model-driven development [17]. The focus on No-code development platform (NCDP) is a related term
the platform is highlighted as a key differentiating aspect used for platforms that eliminate the need for programming
of these solutions with respect to the previous generation of using visual languages, graphical user interfaces, and config-
declarative tools: LCDPs are platforms first, with features for uration. While the term is widely used in marketing, market
application deployment and life-cycle management, as well analysis firms currently oppose using it to identify a clear
as platform management [50]. market segment [47]. For the context of this paper, we con-
Gartner identified a similar segment in 2016, called sider NCDP and LCDP interchangeably, and consequently,
low-code application platform (LCAP) [58]. In particular, hereafter, we use the term LCDP only.
they introduced enterprise LCAPs, which aim at produc-
ing enterprise-class applications requiring high performance,
scalability, high availability, disaster recovery, security,
SLAs, resource use tracking, technical support from the 1 The survey included more than 3 thousand developers in Australia,
provider, and API access to and from local and cloud ser- Canada, China, France, Germany, India, the UK, and the USA. Devel-
vices. opers participated with small material incentives.

123
Low-code development and model-driven engineering 439

Fig. 3 A simple domain model specified in Mendix (from [51])

Fig. 2 Top-level features of LCDPs (refinement of [51])


the Reusability support feature shown in Fig. 2.
For instance, Salesforce App Cloud [52] includes the exten-
3 Overview of low-code development sive AppExchange marketplace [53] consisting of pre-built
platforms applications and components, reusable objects and elements,
drag-and-drop process builder, and inbuilt Kanban boards. As
In this section, we present an overview of the most significant discussed in [51], realizing software systems with LCDPs
LCDPs by considering the typical steps that are performed encompasses several tool-supported steps, which are sum-
when using them and by relying on a refined taxonomy orig- marised in the following.
inally presented in [51].
LCDPs support the development of applications that can Domain modelling In this phase, users are provided with
be web-only or also native for the target deployment envi- modelling constructs to represent concepts and rela-
ronments. Thus, they can natively support both desktop tionships underpinning the application being developed.
and mobile devices and integrate with existing workflows Figure 3 shows a simple domain model specified with
developed with popular Software-as-a-Service (SaaS) appli- the Mendix platform to describe training courses. In par-
cations, including Zapier, Amazon AppFlow, and Trello to ticular, a training Course is defined in terms of its
mention just a few. Appian [2] is among the most long-lived title, description, duration, and price. The data entity
LCDP, whereas Amazon Honeycode [1] and Google App- ScheduledCourse defines when a given course is
Sheet [15] are among the most recent approaches. scheduled in terms of start and end dates. Moreover,
Particular characteristics that distinguish existing LCDPs to schedule a course, it is necessary to specify its
pertain to the user experience of advanced Graphical Trainer and the Location where it is held. The
user interfaces (see Fig. 2) providing tools and wid- Registration entity is for defining registrations of
gets to enable citizen developers to conceive the desired Trainees that want to attend ScheduledCourses
applications. Drag-and-drop facilities, advanced reporting Pre-built templates can be exploited as starting point
features, decision engines for modelling complex logic, and when defining domain models, and interactive applica-
form builders are just examples of functionalities provided tion analytics are provided out-of-the-box. Other tools
in the front end of LCDPs. Moreover, LCDPs can give follow a similar approach. For example, Codebots [7]
the users some Live collaborative development uses UML to specify domain models that are consumed
support to help developers that are geographically dis- to automatically generate target artefacts, including com-
tributed and who want to work on the same applications plete REST APIs, client libraries, Swagger API doc-
collaboratively. Another distinguishing aspect of existing umentation, and a JSON Schema definition for each
LCDPs is related to the Supported application domain object.
domain intended to be the primary focus of interest. For User interface definition Users define data forms and pages
instance, the main focus of Node-RED [34] is supporting the to create, edit, and visualize data that the application
development of IoT applications. Other platforms support under development will manage. Figure 4 shows a form-
the development of chatbots [40], whereas the majority of based screen in Microsoft Power Apps. According to
existing LCDPs aim at being general-purpose supporting the the given specification, the modelled application consists
development of any data-intensive application. of three different screens and one of them is being
LCDPs can provide users with pre-defined artefacts, defined as shown in Fig. 4. The screen under definition
which can be used as starting points. This is reflected by consists of different fields (vertically aligned) that are

123
440 D. Di Ruscio

Fig. 5 Business logic specification with Node-RED [35]

Fig. 4 User interface definition with Microsoft PowerApps [31]

selected from a SharePoint document as shown on the


right-hand side of Fig. 4.
Fig. 6 Configuring the Google Drive connector in Zoho Creator [63]

Business logic specification Users define the control and build mechanisms. Several execution environments
data flows of the system under development through intu- can be supported, as for instance, in the case of zAppDev
itive Business logic specification [61], which provides users with different code generation
mechanisms. Graphical workflows and textual busi- facilities. Once the desired system has been specified and
ness rules are examples of business logic specifications built, a dedicated Deployment support is available
that typically use one or more API call(s). Figure 5 shows to deploy the system in private or public environments.
a simple Node-RED workflow, which retrieves the most Deployments are typically done on cloud infrastructures
recent quakes from an online service and represents the with a few clicks, as shown in Fig. 7. In particular, Out-
retrieved data in a CSV file before performing further System [37] provides developers with quick mechanisms
manipulations. Node-RED implements a programming to publish developed applications, connect different ser-
model that permits developing event-based applications, vices, and create real-time dashboards.
which can be specified by a wide range of node types
available in an extensible palette. Workflow specification
Application maintenance The last step of the process is
is also prominent in Kissflow [24], which mainly focuses
monitoring and maintaining the developed system by
on workflow automation for small businesses.
means of dedicated features, e.g. to react in case of
unforeseen requirements that need to be addressed or fix
Integration with external services LCDPs typically pro- issues that might occur during the operation of the system
vide Interoperability support with external (cf. Monitoring support in Fig. 2).
services and data sources to use services or
consume data provided by third-party systems, e.g. using
dedicated APIs. LCDPs can consume services provided 4 Low-code vs. model-driven engineering
by external providers such as Dropbox, Zapier, Office
365, and Google Drive. Thus, users might connect or Having discussed the main features of LCDPs, we compare
integrate such services to build forms or to compile data them with MDE processes and technologies in this section.
reports. For instance, Fig. 6 shows the page in Zoho Cre- MDE [3] encompasses software paradigms emphasiz-
ator [62] to configure the connection with Google Drive. ing the use of models as first-class artefacts during the
development lifecycle. Hence, in MDE, models are used to
Application generation and deployment The next step of specify, test, simulate, verify, modernize, maintain, under-
the process consists of generating and deploying the mod- stand, and generate code for the system, among many other
elled application by means of provided Application activities. Still, not every MDE process ends with code gen-

123
Low-code development and model-driven engineering 441

approaches termed “model-driven” by our community fall


under regions 1, 2, and 3; with an overlap under 2 and 3 with
low-code platforms and low-code development approaches.
Instead, regions 4 and 5 are exclusively low-code, while
region 1 is exclusively model-driven. The regions can be
described as follows:

Fig. 7 Application deployment with OutSystem [38]


1. This region contains the model-driven approaches that
use models as machine-processable artefacts but do not
model-driven low-code low-code aim at reducing the amount of code required to imple-
engineering application platforms software development ment the system. Instead, they focus on automating
tasks like simulation, formal verification, software opti-
mization, or reverse engineering. Examples of works
in this category include the work of Cortellessa et al.
1 2 [9] on analysing and refactoring UML design models
3 4 5
for optimizing their performance; or reverse engineer-
ing tools like Modisco [5], which extracts models from
code.
2. These are the approaches that use models as machine-
Fig. 8 Venn diagram showing commonalities and differences between processable artefacts and aim to reduce the amount of code
model-driven approaches, low-code application platforms, and low-
code software development
required to implement a system (e.g. via code generation
or interpretation) but without offering deployment or life-
cycle management capabilities for the produced system.
eration but actively uses models. The goal of MDE is to Examples of this class of approaches are JHipster [20] and
increase productivity by automating different steps in soft- its JDL [21] domain-specific language, Google Protocol
ware development employing models while augmenting the Buffers [16], or the OlivaNova model execution system
overall quality [19,23]. For this purpose, MDE processes [39].
often rely on Domain-Specific Languages (DSLs), specially 3. This region contains the platforms that use models to
tailored for the domain at hand. Using domain-specific mod- facilitate the development of software applications with
els makes descriptions more intentional and includes less reduced code and offer built-in deployment and lifecy-
accidental details than code written using general-purpose cle management facilities for the produced application.
programming languages. Hence, those models become eas- Examples include the Codebots [7] and Judo [22] low-
ier to create, verify, and maintain than the corresponding code platforms, both of which are based on technologies
low-level code. of the Eclipse Modelling ecosystem [56].
In their turn, LCDPs promote the construction of appli- 4. This region and the next one contain approaches that
cations using forms and graphical editors with little or no cannot be considered model-driven. In particular, region
hand-crafted code. Since some of their target users are 4 contains the platforms that facilitate the develop-
citizen developers, one of their key points is to reduce acci- ment of software applications with reduced code. Such
dental complexity regarding the installation and operation approaches offer built-in deployment and lifecycle man-
of both the development environments and the developed agement facilities for the produced application. How-
applications. This way, they typically provide cloud-based ever, they do not use models that conform to explic-
development environments and manage the lifecycle of the itly defined languages/metamodels (e.g. they use data
designed applications (e.g. hosting, resource allocation and stored in a relational database or schema-less XML/JSON
provisioning, usage analytics, etc.). Therefore, low-code documents).
development shares some of the goals of MDE, but there 5. These approaches aim to reduce the amount of code
are some differences, too. required to implement a system without offering deploy-
Figure 8 schematically illustrates the commonalities and ment or lifecycle management capabilities for the pro-
differences between low-code and MDE approaches using duced system, and—like region 4—without using models
a Venn diagram. The diagram represents the approaches that conform to explicitly defined languages/metamodels.
following MDE, low-code development, and development Examples of this type of approach include database-
based on low-code platforms in terms of sets. This leads to schema-driven generators like Phreeze [41] and one-off
5 regions of interest (marked as 1–5 in the figure). This way, generators such as those provided by Ruby on Rails [44].

123
442 D. Di Ruscio

Next, we elaborate on other aspects that differentiate 5 Low-code development: why now?
model-based and low-code approaches, based on Fig. 8:
In terms of their core ambition to expedite the delivery of soft-
Platform Low-code application platforms (regions 3 and 4 ware systems, LCDPs are not very dissimilar to previously
in the figure) are mostly cloud-based: they can be used tried approaches like 4GLs, CASE tools, etc., as already men-
from the web browser and host the defined applications. tioned in Sect. 2. Essentially, they provide an environment for
This frees the user from both installing the development specifying the structure and behaviour of a software system
platform itself and from deploying the defined applica- at a high level of abstraction. Such an environment shields
tions. This approach simplifies the adoption of low-code developers from low-level concerns (e.g. specific databases,
by newcomers. While MDE solutions can be cloud-based object-relational mappers, services, messaging, and security
(falling in region 3) [8], this is not the norm today. Instead, middleware). They then generate executable code that real-
many solutions are based on the desktop, for example, izes the specified software system. Given the broad consensus
those using the Eclipse Modelling Framework (EMF) that 4GLs and CASE tools were not wildly successful, why
[56], or meta-modelling tools like MetaEdit+ [30]. These should low-code environments fare any better? There are
approaches would fall under region 2—and may be con- multiple reasons why this may be the case, which we analyse
sidered low-code development approaches—if their aim next.
is automating application development, otherwise they
would fall in region 1. Please note that not all low-code
software development approaches are cloud-based; in Cloud-based deployment Beyond generating code, mod-
particular, those in regions 2 and 5 are not. ern LCDPs can also deploy the produced software sys-
Users LCDPs mainly target end-users, so-called citizen tems on scalable cloud-based infrastructures and make
developers. Therefore low-code platforms tend to be easy them instantly available to users globally through web-
to use for people with a non-technical background. This based interfaces. This can dramatically shorten the time
means that frequently (but not always), users of tools in and effort required to release applications (and updates)
regions 3–4 are citizen developers and non-professional to users and increase the appeal for LCDPs as a medium
programmers. For example, while low-code platforms for rapid application development and delivery.
like OutSystems target citizen developers, others like Digital native workforce Computer literacy has improved
Judo target teams of business analysts, software archi- dramatically over the last 40 years. The basics of com-
tects, and programmers. puter programming are taught in many countries as part
In their turn, MDE solutions can target end-users, but of compulsory education, and the new generations of
many of them are directed to professional software domain experts (e.g. accountants, medicinal practition-
developers since they are expected to be used within ers, construction engineers) are digital natives. As a
development processes. Therefore, typically, users of consequence, while most domain experts would require
approaches in regions 1 and 2 have a more technical back- substantial training to master some part of the complex-
ground. ity of a CASE tool released 40 years ago, a growing
Domains As mentioned in Sect. 2, the first wave of low- number of contemporary domain experts have substantial
code targeted business applications. Recently, we are experience with working with computers and non-trivial
witnessing proposals for low-code tools in other domains, software, and arguably require a lot less training to use a
like IoT/event-driven applications (e.g. Node-RED [34]), LCDPs to implement bespoke applications.
chatbots (e.g. Google’s Dialogflow [10], Amazon’s Lex Zero setup The fact that many LCDPs are cloud-based and
[25], IBM’s Watson Assistant [59]), or Machine Learn- do not require installation of bespoke software signifi-
ing (e.g. Google’s AutoML [14] or RapidMiner [42]). cantly lowers the entry barrier for new users, who can
MDE solutions (in regions 1–3 of the figure) can target evaluate such platforms and even develop and deliver
those domains but frequently also target more technical small-scale applications at no cost from the familiar envi-
areas, which require specialized engineers. These include ronment of their web browser.
domains like automotive [11], power engineering [13], Developer shortfall As software is becoming pervasive
or cyber-physical systems [33] in general, among many in all aspects of human activity, the demand for soft-
others. ware developers has outgrown the supply of suitably
skilled professionals, and the gap is constantly widen-
ing [4]. Moreover, highly skilled software developers
are attracted to intellectually demanding (and financially
rewarding) software systems instead of run-of-the-mill
applications. This creates a growing gap for business

123
Low-code development and model-driven engineering 443

applications that would be more effective than shared of core MDE technologies is underway with frameworks
spreadsheets but are too expensive to implement and such as Xtext [60] and Sirius [55] providing web-based
maintain manually. counterparts. However, significant effort is still required
Training facilities The media through which users learn to realise the vision of zero-installation web-based MDE
have also changed considerably recently. A couple of workbenches. Some efforts already started to reuse open-
decades ago, the primary learning media for application source technologies for building up LCDPs [36]. As
development environments were books written by tech- there is currently already a trend to migrate MDE tech-
nology experts. This landscape has changed dramatically nologies to the web/cloud, there may be an opportunity
with the growth of the web and, particularly, video shar- to develop the next generation of LCDPs with existing
ing services such as YouTube, making it easier to deliver MDE technologies such as metamodelling frameworks
up-to-date training material aimed at different audiences. for language engineering, code generators for produc-
This enables citizen developers to develop and share their ing the final applications, etc. [57]. This may be further
own training material (e.g. walk-throughs, screencasts) supported by current initiatives for building open-source
rather than acting as passive consumers. cloud platforms such as GAIAX [12], which is especially
important for long-living software systems.
Counteracting vendor lock-in Since the introduction of
6 What MDE can learn from low-code and CASE tools, one of the major concerns is the poten-
vice versa tial for vendor lock-in, i.e. application development and
deployment are bound to a particular technology. While
Based on the previously presented insights in low-code this may not be considered as a potential problem in the
development and MDE, we will now discuss what the two short term, it can become critical in the long term. For
approaches can learn from each other to tackle critical chal- instance, consider migrating projects from CASE tools
lenges for their future developments. to MDE tools or projects developed with Rapid Appli-
cation Development (RAD) approaches to modern cloud
platforms. In the context of low-code such issues may
Generic vs. specific platforms Many LCDPs attempt to also occur, e.g. an LCDP that produces applications that
cover a wide range of applications through an ever- only work with a specific cloud provider’s technology
growing library of highly configurable components. In stack (cf. cloud vendor lock-in). Nevertheless, there are
the MDE community, it is widely accepted that in many even more important aspects related to the development
cases, smaller domain-specific languages can be more artefacts. First of all, is an export of the development
beneficial for engagement with domain experts and auto- artefacts possible, and if it is, how can these artefacts be
mated reasoning and processing than large and complex reused, imported, and interpreted in other platforms? The
all-encompassing languages such as UML. An open MDE community has invested substantial effort in this
question is if the current generation of domain-agnostic respect by providing dedicated standards for modelling
LCDPs will increasingly struggle as they keep growing in languages (e.g. UML, BPMN), and even meta-modelling
complexity. This can give rise to domain-specific LCDPs languages (e.g. MOF, Ecore), model exchange standards
in the future, which will target specific classes of sys- (e.g. XMI and HUTN), etc. It has to be explored if these
tems and citizen developers. Here an opportunity is about approaches may also be reused for LCDPs or if other
reusing the rich technological infrastructure offered by means are needed to prevent vendor lock-in.
MDE for building domain-specific platforms. Interest- Fostering ecosystems Providing an LCDP is the first step,
ingly, while MDE is often referred to as an essential but then an ecosystem for this platform is required to
building block of low-code in the Forrester and Gartner ensure the continuous growth of a healthy user base. This
reports, there is little evidence that existing LCDPs use may be even more important for LCDPs as professional
technologies (predominately Eclipse-based) commonly developers as well as citizen developers may be targeted.
used in the MDE community. Thus, it seems to be an Thus, the availability of documentation, support, consul-
opportunity to speed up the development of LCDPs with tancy, reusable components, etc., is of major importance.
MDE technologies if the latter are ready to run on the In MDE, such an ecosystem was triggered by Eclipse, i.e.
web/cloud and can deal with the requirements of typical a large and active ecosystem around the Eclipse Mod-
LCDP users. elling Framework was established from the industrial
Opening up web/cloud-based platforms A lesson that the and academic sides. It has to be further explored how
MDE community can learn from the success of LCDPs such ecosystems will develop for LCDPs, as most cur-
is that web-based interfaces can significantly improve rent platforms are single vendor efforts. This issue also
uptake and engagement with domain experts. A transition concerns the academic area, where scientific community

123
444 D. Di Ruscio

efforts are required to stimulate research on topics related ment no. 813884. The work has also been partially funded by the Spanish
to low-code [57]. For instance, a current example is the Ministry of Science (RTI2018-095255-B-I00) and the R&D programme
of Madrid (P2018/TCS-4314).
low-code workshop [27] hosted with the MODELS con-
ference since 2020, which provides a forum to discuss Open Access This article is licensed under a Creative Commons
low-code development and MDE. Attribution 4.0 International License, which permits use, sharing, adap-
Managing software evolution Notably, one of the most tation, distribution and reproduction in any medium or format, as
long as you give appropriate credit to the original author(s) and the
crucial stages of the software lifecycle is the maintenance source, provide a link to the Creative Commons licence, and indi-
of a software product after its release. Providing support cate if changes were made. The images or other third party material
to such activities requires the ability to grow in function- in this article are included in the article’s Creative Commons licence,
ality and size without unwanted side effects satisfying unless indicated otherwise in a credit line to the material. If material
is not included in the article’s Creative Commons licence and your
new requirements emerging from the routine usage of intended use is not permitted by statutory regulation or exceeds the
the product. Managing software evolution processes in permitted use, you will need to obtain permission directly from the copy-
LCDPs is an interesting line of research since these plat- right holder. To view a copy of this licence, visit http://creativecomm
forms are managed and allow cloud-based monitoring ons.org/licenses/by/4.0/.
of the developed applications. Consequently, the plat-
form provider should offer as much support for evolution
as possible. However, this may involve many different
aspects. Considering the application level, we may need References
support for model/data co-evolution, e.g. the data model
1. Amazon Honeycode. https://www.honeycode.aws/. Accessed Sept
is changing and there are already running instances of 2021
the application in usage. Evolution also applies on the 2. Appian. https://appian.com/. Accessed Sept 2021
language level, which has been extensively researched 3. Brambilla, M., Cabot, J., Wimmer, M.: Model-Driven Software
Engineering in Practice. Synthesis Lectures on Software Engineer-
in MDE, and is often referred to metamodel/model
ing, 2nd edn. Morgan & Claypool Publishers, San Rafael (2017)
co-evolution [18]. Here, the problem applies both to 4. Breaux, T., Moritz, J.: The 2021 software developer shortage is
low-code and MDE approaches. Low-code will only coming. Commun. ACM 64(7), 39–41 (2021)
be successful if applications developed with low-code 5. Brunelière, H., Cabot, J., Dupé, G., Madiot, F.: Modisco: a model
driven reverse engineering framework. Inf. Softw. Technol. 56(8),
approaches can evolve for a longer time in combination
1012–1032 (2014)
with the LCDPs themselves. 6. Cabot, J.: Positioning of the low-code movement within the field
of model-driven engineering. In: Guerra, E., Iovino, L. (eds.)
MODELS ’20: ACM/IEEE 23rd International Conference on
Model Driven Engineering Languages and Systems, Virtual Event,
7 Summary Canada, 18–23 October, 2020, Companion Proceedings, pp. 76:1–
76:3. ACM (2020)
7. Codebots. https://codebots.com/. Accessed Sept 2021
This paper compared and positioned the relatively new 8. Corley, J., Syriani, E., Ergin, H.: Evaluating the cloud architecture
low-code movement against the established model-driven of AToMPM. In: Proceedings of MODELSWARD, pp. 339–346.
engineering discipline. We summarised the history of low- SciTePress (2016)
9. Cortellessa, V., Eramo, R., Tucci, M.: From software architecture
code so far, provided an overview of typical low-code to analysis models and back: model-driven refactoring aimed at
development processes and the tools that LCDPs offer to availability improvement. Inf. Softw. Technol. 127, 106362 (2020)
support them, and contrasted and compared the principles 10. Dialogflow. https://dialogflow.com/. Accessed Sept 2021
and practices of low-code and model-driven engineering. 11. Drave, I., Hillemacher, S., Greifenberg, T., Kriebel, S., Kusmenko,
E., Markthaler, M., Orth, P., Salman, K.S., Richenhagen, J., Rumpe,
While low-code and model-driven engineering both aspire B., Schulze, C., von Wenckstern, M., Wortmann, A.: SMArDT
to improve software development by raising abstraction and modeling for automotive software testing. Softw. Pract. Exp. 49(2),
hiding implementation-level details, we argue that the two 301–328 (2019)
practices are not identical. Indeed, not all model-driven 12. GAIAX. https://www.data-infrastructure.eu/GAIAX/Navigation/
EN/Home/home.html. Accessed Sept (2021)
approaches aim at reducing the amount of code needed 13. Gómez, A., Mendialdua, X., Barmpis, K., Bergmann, G., Cabot,
to implement software solutions, and not all low-code J., Carlos, X.D., Debreceni, C., Garmendia, A., Kolovos, D.S., de
approaches are model-driven. However, being close con- Lara, J.: Scalable modeling technologies in the wild: an experience
ceptually creates substantial potential for applying existing report on wind turbines control applications development. Softw.
Syst. Model. 19(5), 1229–1261 (2020)
knowledge and cross-pollination between the two disci- 14. Google. AutoML. https://cloud.google.com/automl/. Accessed
plines. Sept 2021
15. Google AppSheet. https://www.appsheet.com/. Accessed Sept
Acknowledgements This work has received funding from the Low- 2021
comote project under European Union’s Horizon 2020 research and 16. Google’s protocol buffers. https://developers.google.com/
innovation program under the Marie Skłodowska-Curie grant agree- protocol-buffers. Accessed Sept 2021

123
Low-code development and model-driven engineering 445

17. Hammond, J.: The Forrester Wave: Mobile Low-Code Develop- 47. Rymer, J., Koplowitz, R.: Now Tech: Rapid App Delivery, Q1 2019.
ment Platforms, Q1 2017. Forrester Research, Cambridge (2016) Forrester Research, Cambridge (2019)
18. Hebig, R., Khelladi, D.E., Bendraou, R.: Approaches to co- 48. Rymer, J., Koplowitz, R.: The Forrester Wave: Low-Code Devel-
evolution of metamodels and models: a survey. IEEE Trans. Softw. opment Platforms For AD&D Professionals, Q1 2019. Forrester
Eng. 43(5), 396–414 (2017) Research, Cambridge (2019)
19. Hutchinson, J.E., Whittle, J., Rouncefield, M.: Model-driven engi- 49. Rymer, J., Richardson, C.: The Forrester Wave: Low-Code Devel-
neering practices in industry: social, organizational and managerial opment Platforms, Q2 2016. Forrester Research, Cambridge (2016)
factors that lead to success or failure. Sci. Comput. Program. 89, 50. Rymer, J., Richardson, C.: Vendor Landscape: The Fractured.
144–161 (2014) Fertile Terrain of Low-Code Application Platforms. Forrester
20. JHipster. https://www.jhipster.tech. Accessed Sept 2021 Research, Cambridge (2016)
21. JHipster’s JDL DSL. https://www.jhipster.tech/jdl. Access Sept 51. Sahay, A., Indamutsa, A., Ruscio, D.D., Pierantonio, A.: Support-
2021 ing the understanding and comparison of low-code development
22. Judo. https://judo.codes. Accessed Sept 2021 platforms. In: Proceedings of 46th Euromicro Conference on Soft-
23. Kelly, S., Tolvanen, J.: Domain-Specific Modeling—Enabling Full ware Engineering and Advanced Applications SEAA, pp. 171–178.
Code Generation. Wiley, Hoboken (2008) IEEE (2020)
24. Kissflow. https://kissflow.com/workflow/process/. Accessed Sept 52. Salesforce. https://developer.salesforce.com/. Accessed Sept 2021
(2021) 53. Salesforce (AppExchange marketplace). https://appexchange.
25. Lex. https://aws.amazon.com/en/lex/. Accessed Sept 2021 salesforce.com/. Last Accessed Sept 2021
26. Lieberman, H., Paternò, F., Klann, M., Wulf, V.: End-User Devel- 54. Schmidt, D.: Guest editor’s introduction: model-driven engineer-
opment: An Emerging Paradigm, pp. 1–8. Springer, Dordrecht ing. Computer 39, 25–31 (2006)
(2006) 55. Sirius. https://www.eclipse.org/sirius/. Accessed Sept 2021
27. LowCode Workshop at MODELS. https://lowcode-workshop. 56. Steinberg, D., Budinsky, F., Paternostro, M., Merks, E.: EMF:
github.io/. Accessed Sept 2021 Eclipse Modeling Framework, 2nd edn. Addison-Wesley Profes-
28. Martin, J.: Application Development Without Programmers. Pren- sional (2008). see also https://www.eclipse.org/modeling/
tice Hall PTR, Hoboken (1982) 57. Tisi, M., Mottu, J., Kolovos, D.S., de Lara, J., Guerra, E., Rus-
29. Martin, J.: Rapid Appl. Dev. Macmillan Publishing Co. Inc, New cio, D.D., Pierantonio, A., Wimmer, M.: Lowcomote: training
York (1991) the next generation of experts in scalable low-code engineering
30. MetaEdit+ by Metacase. https://www.metacase.com/products. platforms. In: STAF 2019 Co-Located Events Joint Proceedings,
html. Accessed Sept 2021 volume 2405 of CEUR Workshop Proceedings, pp. 73–78. CEUR-
31. Microsoft Power Apps. https://docs.microsoft.com/en-us/ WS.org (2019)
powerapps/maker/canvas-apps/working-with-forms. Accessed 58. Vincent, P., Iijima, K., Driver, M., Jason, W., Natis, Y.: Magic
Sept 2021 Quadrant for Enterprise Low-Code Application Platforms. Gartner
32. Mines, C.: Predictions 2020: More Changes for Software Devel- (2016)
opment. Forrester Research, Cambridge (2020) 59. Watson. https://www.ibm.com/cloud/watson-assistant/. Accessed
33. Mohamed, M.A., Kardas, G., Challenger, M.: Model-driven engi- Sept 2021
neering tools and languages for cyber-physical systems. A system- 60. Xtext. https://www.eclipse.org/Xtext/. Accessed Sept 2021
atic literature review. IEEE Access 9, 48605–48630 (2021) 61. zAppDev. https://zappdev.com/. Accessed in Sept 2021
34. Node-RED. https://nodered.org/. Accessed 2021 62. Zoho Creator. https://www.zoho.com/creator/. Accessed Sept
35. Node-RED (workflows). https://nodered.org/docs/tutorials/ 2021
second-flow. Accessed Sept 2021 63. Zoho Creator (third-party integration). https://www.zoho.com/
36. OSBP. https://www.eclipse.org/osbp/. Accessed Sept 2021 developer/help/extensions/connectors.html. Accessed Sept 2021
37. Outsystems. https://www.outsystems.com/. Accessed Sept 2021
38. Outsystems (deploying an application). https://success.outsystems.
com/Documentation/11/Managing_the_Applications_Lifecycle/
Deploy_Applications/Deploy_an_Application. Accessed Sept Publisher’s Note Springer Nature remains neutral with regard to juris-
2021 dictional claims in published maps and institutional affiliations.
39. Pastor, O., Gómez, J., Insfrán, E., Pelechano, V.: The OO-method
approach for information systems modeling: from object-oriented
conceptual modeling to automated programming. Inf. Syst. 26(7), Davide Di Ruscio is an Asso-
507–534 (2001) ciate Professor at the Università
40. Pérez-Soler, S., Juarez-Puerta, S., Guerra, E., de Lara, J.: Choosing degli Studi dell’Aquila (Italy). His
a chatbot development tool. IEEE Softw. 38(4), 94–103 (2021) main research interests are related
41. Phreeze. http://www.phreeze.com. Accessed Sept 2021 to several aspects of Software
42. RapidMiner. https://rapidminer.com/. Accessed Sept 2021 Engineering, Open-Source Soft-
43. Richardson, C., Rymer, J.: New Development Platforms Emerge ware, Model-Driven Engineering
for Customer-Facing Applications. Forrester Research, Cambridge and Recommender Systems. He
(2014) has published more than 170
44. Ruby, S., Copeland, D., Thomas, D.: Agile Web Develop- papers in various journals, con-
ment with Rails 6. The Pragmatic Programmers, 2019. See ferences, and workshops on such
also https://guides.rubyonrails.org/command_line.html#bin-rails- topics. He is in the editorial board
generate. Accessed Sept 2021 of the International Journal on
45. Rymer, J.: The Forrester Wave: Low-Code Development Platforms Software and Systems Modeling
For AD&D Pros, Q4 2017. Forrester Research, Cambridge (2016) (SoSyM), of IEEE Software, of
46. Rymer, J.: Siemens Snaps Up Mendix; Low-Code Platforms Enter the Journal of Object Technology, and of the IET Software journal.
New Phase. Forrester Research, Cambridge (2018) Contact him at [email protected] or visit http://people.disim.
univaq.it/diruscio/.

123
446 D. Di Ruscio

Dimitris Kolovos is a Professor Massimo Tisi is an associate


of Software Engineering in the professor in the Department of
Department of Computer Science Computer Science of the Institut
at the University of York, where Mines-Telecom Atlantique (IMT
he researches and teaches auto- Atlantique, Nantes, France), and
mated and model-driven software deputy leader of the NaoMod
engineering. He is also an Eclipse team, LS2N (UMR CNRS 6004).
Foundation committer, leading the Since 2019 he coordinates the
development of the open-source Lowcomote Marie Curie European
Epsilon model-driven software Training Network. He has been
engineering platform, and an edi- visiting researcher at McGill Uni-
tor of the Software and Systems versity and the National Institute
Modelling journal. He has of Informatics (NII) in Japan and
co-authored more than 150 peer- post-doctoral fellow at Inria. He
reviewed papers, and his research received his PhD degree in Infor-
has been supported by the European Commission, UK’s Engineering mation Engineering at Politecnico di Milano (Italy), where he was a
and Physical Sciences Research Council (EPSRC), InnovateUK, and member of the Database and Web Technologies group. His research
by companies such as Rolls-Royce and IBM. interests revolve around software and system modeling, domain-
specific languages, and applied logic. He contributes to the design of
Juan de Lara is full professor at the ATL model-transformation language and investigates the applica-
the computer science department tion of deductive verification techniques to model-driven engineering.
of the Universidad Autónoma de
Madrid. Together with Esther Manuel Wimmer is full professor
Guerra, he leads the modelling leading the Institute of Business
and software engineering research Informatics—Software Engineer-
group. His research interests are ing at the Johannes Kepler Uni-
in model-driven engineering and versity Linz, and he is the head
automated software development. of the Christian Doppler Labo-
Contact him at juan.delara@ ratory CDL-MINT. His research
uam.es, or visit http://arantxa.ii. interests comprise foundations of
uam.es/~jlara/. model engineering techniques as
well as their application in
domains such as tool interoper-
ability, legacy modeling tool mod-
ernization, model versioning and
Alfonso Pierantonio is full pro- evolution, and industrial engineer-
fessor at the Università degli Studi ing. For more information, please
dell’Aquila (Italy). His interests visit https://www.se.jku.at/manuel-wimmer.
are in software engineering,
model-driven engineering,
and language engineering with spe-
cial attention to co-evolution tech-
niques, consistency management,
and bidirectionality. He has pub-
lished more than 160 articles in
scientific journals and conferences
and has been on the organizing
committee of several international
conferences, including MoDELS
and STAF. Alfonso is Editor-in-
Chief of the Journal of Object Technology and in the editorial and
advisory board of Software and System Modeling, and Science of
Computer Programming. He has been PC Chair of ECMFA 2018,
General Chair of STAF 2015, and is a Steering Committee mem-
ber of the ACM/IEEE MoDELS. He is a co-principal investiga-
tor of several research and industrial projects. You can contact the
author at [email protected] or visit http://disim.univaq.it/
AlfonsoPierantonio

123

You might also like