Towards a Multi-Leveled Architecture

for the Internet of Things

Rayhana Bouali Baghli∗† , Elie Najm∗ and Bruno Traverson†
∗ LTCI,CNRS, Télécom ParisTech, Université Paris Saclay
46 rue Barrault, 75013, Paris, France
Email: {rayhana.baghli, elie.najm}@telecom-paristech.fr
† EDF R&D, Département ICAME,

7 Boulevard Gaspard Monge, 91120 Palaiseau

Email: [email protected]

Abstract—We propose a data-centric three leveled modeling For our research, we consider objects from the perspective
architecture in an effort towards a Model Driven approach of of their uses. An object can be used in multiple applications.
services for the Internet of Things (IoT): a resources level, an As part of the establishment of a security service, one or
artifacts level and a semantic level. In this architecture, the
resources level abstracts all important pieces of information more presence sensors are used to detect movement and react
describing real objects as resources. The artifacts level allows accordingly, triggering an alarm, for example. An automatic
to collect all objects and contexts information necessary for lighting service also uses a presence sensor for detecting the
the execution of a given service. The semantic level introduces movement and turn on or off the light according to presence or
semantic notions to the architecture. So, data and actions are absence of movements. Today, in IoT, there is still no standard
named in a standardized naming and the rules facilitate the
interaction of the system with the non-expert users. We illustrate framework approved and adopted by the scientific community
our architecture on a small example in which we present all three and by manufacturers of connected objects. Instead, the trend
levels. is to produce objects that communicate via proprietary models
and protocols, which are based on vertical architectures, lead-
I. I NTRODUCTION
ing to the emergence of technology silos. In the current model
In recent years, the development of so-called smart objects in silos, a security service and an automatic lighting services
has gained relevance i.e objects with capacity of communica- can not share the same presence sensor if they are not from
tion and execution. Individually, each of these objects offers the same manufacturer, but each service has its own presence
one or several basic services, but all these smart objects are sensor as shown in Figure 2.
also a promise for the development of richer services. This
phenomenon is called the Internet of Thing (IoT). The IoT
is changing market dynamics and should have a significant
impact on the nature of products and services provided to
users and therefore, the overall model architecture. [5]
Figure 1 shows a sample of smart objects. A smart object
could be a temperature sensor, a connected car or also a
Fig. 2. Redundancy of a presence sensor in vertical architectures
connected fridge. Objects in the IoT have very different
characteristics. Some objects are mobile and others not. Some The example silos shown in Figure 2 is due to the strong
objects could measure physical properties and others not. coupling between objects and services that rely on these
Some objects could be remotely activated. Some object can objects. One of the major challenges of our study is to propose
execute programs and also take decisions locally. Others don’t a horizontal architecture that decouples the connected objects
and could only execute an order received from the network. from available services. Then, several services could access to
the same object if they need. Thus, the lighting service and
the security service can share the same presence capture data
returned by the presence sensor as shown in Figure 3.

Fig. 1. Sample of connected objects.

One of the major research challenges in the IoT is to control

the heterogeneity and find a way to represent homogeneously
all objects that constitute the Internet of Things. Fig. 3. Sharing a presence sensor in a horizontal architecture

978-1-4673-9933-3/16/$31.00 ©2016 IEEE

182
 Beyond interoperability between objects, the second impor- introduction. After that, we describe linking between the three
tant point of our research is to weave these objects together in level of the architecture.
order to provide new services to users. We have the ambition
to dynamically compose services offered by the objects to A. Resources Level
produce new and more complex services and to make them The first level of our architecture is a resources meta-model.
available on the network. These new services can then be This meta-model aims to address the issue of heterogeneity of
reused and also be composed to create other new services. objects and their communications in the Internet of Things.
The third important point of our research is the accessibility The resources meta-model is based on the REST (Represen-
of service composition rules to end users. in the IoT, the tational State Transfer) architectural style. REST has been
service composition process and the rules thereunder are often proposed by Roy T. Fielding in 2000 during his doctoral thesis
intended to satisfy a need or a request of the user. This request and is intended for development of distributed applications
is made declaratively describing objectives. However, rules of [11]. It is based on creation of loosely coupled services that
service composition are often written in imperative languages can easily be reused.
that describe a procedure rather than a goal. These languages In this meta-model, regardless of the connected object, its
are often too technical for non-experts. Our work tends to bring type and its characteristics, the object is represented as a
these two approaches so that users can initially approve the resource. Resources are addressable and available online. The
rules proposed to them by composer of services and secondly resources meta-model allows to homogenize nature of objects
complete or change those composition rules. by encapsulating them in a resource. This meta-model also
To address these issues, we propose a model of architecture provides a uniform access to resources. Unlike architectures
in three levels. The first level, called resources level, is based based on SOA (Service Oriented Architectures) [10] where
on a ROA (Resource Oriented Architecture) model [15] and methods of access to services differ from one service to an-
responds to the problem of heterogeneity, interoperability and other, architectures based resources ROA (Resource Oriented
communication between connected objects. The second level, Architectures) standardize access to resources [15]. Resources
called artifacts level, is an abstraction of resources and can are then accessible in a uniform way. Whatever the resource on
offer a view of operational services. It helps to break current the Web, it is always queried using the same methods CRUD
technology silos of service composition in the IoT. Finally, (Create, Retrieve, Update and Delete).
the third level, called semantic level adds meta contextual data Architectures that follow REST principles are called Restful
to data sent over the network. The use of these meta data in architectures. In this type of architecture, interaction between
service composition process makes this process more dynamic. resources are made by Request/Response mechanism [11].
Meta data are also used to interface with non-expert users in To complement these exchanges with asynchronous event-
order to make composition process and rules accessible for exchange that are essential in the context of the IoT, Re-
them. quest/Response mechanism is enriched by Publish/Subscribe
The application context of our study is energy management mechanism. Publish/Subscribe allows to subscribe to state of a
in smart homes. We focus on the connected objects that are resource and to be informed by notification when the resource
present in smart homes and we consider that each connected state changes without requesting again.
object provides to its environment a basic service. The idea
is to weave the various objects in order to reuse the services
they offer and compose them in order to produce new richer
services.
The remainder of this article is organized as follows. In
section 2, we describe our proposed architecture by defining
its three levels and their links. We show, in section 3, an
application example of the proposed architecture in the context
of the smart home. In section 4, we analyze the relevant
research and technology contexts around interoperability and
collaboration in the field of the IoT. We discuss, in section 5,
our proposal and then we conclude by presenting perspectives
to our work.

II. M ULTI - LEVELED A RCHITECTURE

Fig. 4. Resources Meta-Model
We present in this chapter, our proposed architecture that
features three levels : resources level, artifacts level and Figure 4 shows the resources meta-model of our architec-
semantic level. We describe each level in a subsequent sub- ture. In this meta-model, each connected object is represented
section in which we present the technologies involved and by a resource. At a given time, an object is located in
the issues they solve among those we have explained in the a place that is described by the currentPlace attribute. An

183
object can be a sensor, an actuator, or both simultaneously. that runs to achieve a goal. Each Service describes a set of
Depending on its type, an object can produce data for its DataTypes and ActionTypes it needs to run. Service may have
environment if it is a sensor. It can provide a way to act multiple operating Modes and runs in a given context. In the
on its environment via actions if it is an actuator. Data artifacts meta-model, an Artifact is created for each Service.
and actions that are provided by the connected object are The Artifact has the function of collecting the contextual
defined by DataType and ActionType. On the resources meta- informations for the Service. It also collects DataTypes and
model, each Resource that represents a Connected Object is ActionTypes that are made available by connected objects
characterized by a set of DataTypes and ActionTypes. The and whose Service needs to run. The Artifact has access to
Catalog defines all DataTypes and ActionTypes that can be different operating Modes that the Service offers. The Artifact
provided by Connected Objects. In the context of the smart is described by an Information Model and a Life Cycle. The
home, the HomeRepository describes the architectural plan of Information Model defines the structure of the Artifact and
the house and all its Places. It also describes all Connected the Life Cycle specifies the evolution of the Artifact as for the
Objects that are present in the house. It depicts inhabitants of execution of the Service.
the house and their guests. It finally lists all running Services
in the house. The Place is a geographical location in which C. Semantic Level
can be positioned one or more Connected Objects. The Place The third level of our architecture is built around a semantic
can be either a Room or a Wall. meta-model. This level helps to enrich the architecture with
semantic properties. The semantic meta-model is based on
B. Artifacts Level Semantic Web technologies from work of the World Wild Web
The second level of our architecture is built around a Consortium (W3C). Specifically, Linked Data are intended to
meta-model of artifacts. The artifacts meta-model is based on create a Web of Data or so-called Web 3.0 [9]. To build the
Business Artefacts that are an emerging approach proposed by Web of Data, it is important that the data is available on
IBM in the late 2010s for modeling business processes in a the web in a standard format, accessible and manageable by
company [7] [8]. humans and by semantic web tools.
The Business Artifacts approach aims to combine aspects The semantic meta-model answers to two of the issues
of data and processes in one basic building block, this block described in the introduction. First, in terms of composition of
is called ”Artifact”. The key business entities for the business services, the semantic meta-model aims to make composition
processes are modeled as artifacts. These entities evolve as process more dynamic. It aligns naming of data in accordance
business operations perform. with standardized business vocabularies. Thus, data is typed
The artifacts meta-model offers a horizontal architecture and in a standardized manner according to the kind of information
breaks service composition vertical silos that are currently in it represents. For example, temperature representative data is
place. The artifacts meta-model allows multiple services to always designated by the same type ”temperature”. It is then
access the same objects and the data they generate. easier to replace, in composition process, a missing data with
The artifacts meta-model allows to abstract the resources another data of the same type and has the same characteristics.
meta-model in order to have a global and particular view. Thus, It is also easier to integrate new data into the process of
each service will have a view of resources that corresponds to composition, and thus, new objects, when their types are
its needs. In the context of smart home, instead of having an already known by the composer of services. Actions are typed
overview of all the resources that describe all the objects, the in a standard way too. So, service composer is interested only
artifacts meta-model takes care to select only resources that in types of actions that are provided, disregarding the type
are useful to each service. Each service has a restricted view of object that provides this action. for example, a thermostat
of the resources in the house but different views may intersect service will be interested in the availability of the action of
and more services can have access to the same resources. ”heat” disregarding the type of device that offers this action
(electric radiator, underfloor heating,... etc.). This approach
tends to simplify and make more dynamic composition process
by standardizing data types and action types that are offered
by smart things.
Secondly, data types and action types are written in a stan-
dardized way that is closed to natural language. Business rules
and business policies are built based on these vocabularies. So,
the semantic meta-model allows to write business rules and
policies in a language close to natural language. Therefore, it
bridges the artifacts meta-model which is written in technical
Fig. 5. Artifacts Meta-Model language by non-expert users.
Figure 6 shows the semantic meta-model of our architecture.
Figure 5 shows the artifacts meta-model of our architecture. In this meta-model, Service represents a software package
In this meta-model, Service represents a software package that runs to achieve a goal. To achieve its objective, the

184
 the object offers. This description is transformed into views
present in all three levels.
We described the way that the three levels of our architec-
ture will be joined. It is still an ongoing work and we need to
define more formally the links between each level.

III. A PPLICATION TO SMART HOME

In this section, we present an illustrative example of a
thermostat service in the context of a smart home. We present
the thermostat service and then we will give an example of
concepts for each level of our architecture.

A. illustrative example
We consider a thermostat service that controls the heating
and runs in one room of a smart home. The room includes
Fig. 6. Semantic Meta-Model a heater, a window, an internal thermometer and an external
thermometer.
∙ Internal thermometer gives internal temperature ”IT”;
Service described one or more BusinessProcesses and each
∙ External thermometer gives external temperature ”ET”;
BusinessProcess can support BusinessRules. Each Service
∙ Heater heats the room when turned on, it has three
describes a set of data types and action types it needs to run.
possible states ”off”, ”low” and ”high”. Off state does
A Service may have multiple operating Modes and runs in
not heat and does not consume energy. Low state heats
a given Context. Each Mode is a set of BusinessRules that
with low power and low power consumption. High state
describe how the Service behaves. Context is a complex entity
heats with high power and high power consumption;
that includes all contextual data that the Service needs to
∙ Window has ”open” and ”closed” states.
run. The content of Context entity will be described in detail
in the artifacts level. Catalog describes standardized business The thermostat service offers two operating modes, ”econ-
vocabularies and all informations that are in the Context will omy” and ”comfort” modes. The ”economy” mode favors
be named according to that Catalog. The interaction between energy saving in its temperature control rules and ”comfort”
different Services is managed by a Policy which is a set of mode favors comfort of the inhabitants. User specifies a target
WeavingRules. temperature (TT) and choses the operating mode.
To show the feasibility of the proposed approach, we
D. Linking between Levels chose an algorithm for temperature control which is shown
We presented the three levels of our architecture, we will graphically in Figure 7. This thermostat service is intended for
now describe briefly how we define the links between these use in winter. We then assume that the external temperature
three levels. In fact, the resources, the artifacts and the seman- (ET) is low, the internal temperature (IT) is higher than the
tic levels show different views of the system. This means that external temperature but lower than the temperature desired
the same concept can have complementary views in different by the user (TT) i.e: ET ≤ IT ≤ TT.
levels of the architecture. However, one concept can be partly We consider for each mode a coefficient (𝑅𝑚𝑜𝑑𝑒 ) and a
obtained by transformation from some level to some other tolerance (Δ𝑚𝑜𝑑𝑒 ).
level. For example, the concept of ”Service” is defined at the The interval between ET and TT can be divided into three
semantic level and is visible from the artifacts and resources intervals:
levels. The concepts of ”DataType” and ”ActionType” are ∙ [𝐸𝑇, 𝐸𝑇 + 𝑅𝑚𝑜𝑑𝑒 × (𝑇 𝑇 − 𝐸𝑇 )];

defined in the resources level and are visible from the artifacts ∙ [𝐸𝑇 + 𝑅𝑚𝑜𝑑𝑒 × (𝑇 𝑇 − 𝐸𝑇 ), 𝑇 𝑇 − Δ𝑚𝑜𝑑𝑒 ];

level. ∙ [𝑇 𝑇 − Δ𝑚𝑜𝑑𝑒 , 𝑇 𝑇 ].

We adopted an architectural approach by views but our When IT is in the first interval, the heater is turned on in
approach do not exclude the fact that there are concepts which high position. When IT is in the second interval, the heater
arise from other concepts. For example, in the artifacts level, is turned on in low position; The third interval is an interval
the information model is defined from the definition of the of tolerance, when TI is in this interval, we consider that the
service at the semantic level. target temperature has been sufficiently approached, the heater
for example, smart objects providers, that aim at having their is turned off.
objects be part of a smart home application, should provide, In comfort mode, the heater is left on high mode longer
associated with their objects, a description that includes, the than in economy mode for the internal temperature reaches
objects capabilities, the form of messages exchanges, the faster the target temperature, so 𝑅𝑒𝑐𝑜𝑛𝑜𝑚𝑦 < 𝑅𝑐𝑜𝑚𝑓 𝑜𝑟𝑡 .
objects life cycle, the communication and application protocols The tolerance range is smaller in the comfort mode than in
that it can perform and the data types and action types that economy mode to ensure that IT is closest to TT. In contrast,

185
 that interest the thermostat service. It has data coming from
the internal thermometer, from the external thermometer, from
the heater and and from the window. The artifact is also aware
of different service modes that are available. Artifact make
decisions according to rules described in the semantic meta-
model.
For illustration, we consider the situation where external
temperature is 14∘ C, internal temperature is 17∘ C and target
Fig. 7. Temperature Control Algorithm
temperature is set to 20 ∘ C. The window is closed, the ther-
mostat service is active and the selected mode is the economy
when the economy mode is selected by the user, energy saving mode. According to the temperature control algorithm and the
is more favored than comfort mode, the tolerance range is coefficient and tolerance defined for the economy mode, IT is
bigger to encourage energy saving, i.e: Δ𝑐𝑜𝑚𝑓 𝑜𝑟𝑡 < Δ𝑒𝑐𝑜𝑛𝑜𝑚𝑦 . located on the first interval [ET , ET+ R𝑒𝑐𝑜𝑛𝑜𝑚𝑦 ×(TT-ET)],
∙ Economy Mode the heater is turned on high position.
– R𝑒𝑐𝑜𝑛𝑜𝑚𝑦 = 13
– Δ𝑒𝑐𝑜𝑛𝑜𝑚𝑦 = 1∘ C
∙ Comfort Mode
– R𝑐𝑜𝑚𝑓 𝑜𝑟𝑡 = 23
– Δ𝑐𝑜𝑚𝑓 𝑜𝑟𝑡 = 0.5∘ C
Now, we turn to describe our service in each of the 3 levels.
B. Semantic Level
The semantic level describes the temperature control algo-
rithm in terms of rules that are understandable by non-expert
users. These rules are described in a language close to natural Fig. 8. Example of Artifact state Evolution
language. SBVR (Semantics Of Business Vocabulary And
Rules) [17] is a standard of the OMG (Object Management Figure 8 shows evolution of the artifact state due to a
Group), to serve as a basis for a declarative, formal and change in the internal temperature which takes the value of
executable description of a complex entity in natural language. 18.5∘ C. This change is indicated by an incoming event to
Thus, it represents the first specification of the OMG that the artifact from the resource level. Following this event, the
includes the official use of natural language modeling. We artifact updates the value of the internal temperature. This new
have already studied and used this standard to make business value of IT is in the second interval [ET+R𝑚𝑜𝑑𝑒 ×(TT-ET),TT-
rules understandable for non-technical experts [6]. We propose Δ𝑚𝑜𝑑𝑒 ]. The artifact then redetermines its new state based on
to adopt SBVR to express business rules in a language the new value of IT. This change implies a modification of the
understandable by non-expert users. heater’s power to low. This change is performed on the artifact.
SBVR allows describing business vocabularies and business Hence, a signal is sent from the artifact to the resource level
rules. It describes the business vocabulary by defining noun to change the heater’s power from high to low.
concepts and verbs concepts. Noun concepts can represent D. Resources Level
entities or attributes and verbs concepts can represent actions
or operations. SBVR describes also terms that allow to build This meta-model level is used to represent real-world ob-
business rules from the business vocabulary. SBVR terms can jects as resources. This meta-model joins the artifact and the
for example be modal rules (eg it is mandatory, it is possible), connected objects. The resource meta-model guarantees con-
comparators (eg greater than, lower than), or expression of sistency of communications between resources using CRUD
cardinality (eg exactly one, at least one). Business rules are methods. When the artifact decides to change an object state,
then defined based on the business vocabulary and the SBVR it notifies the resource level, and therefore the resource which
terms. represents the object in question passes this change onto the
Lets see two examples of regulation control rules expressed physical object.
in SBVR. SBVR terms are represented in bold, verb concepts The thermostat service example is a use case that illustrates
are represented in italic, and noun concepts are underlined. the modeling of the three-levels architecture that we proposed.
∙ It is possible that heating performs on more than one
It is still a case of simple use of temperature control but which
heater. illustrates the feasibility of our approach.
∙ It is mandatory to turn heating on if window is open
IV. R ELATED W ORK
C. Artifacts Level Architectures based on resources are very successful among
The architecture defines one artifact per service. So, there communities working around communications in the Internet
is one thermostat artifact and it has visibility of all objects of Things. This success is due in particular to the low coupling

186
properties, simplicity and scalability of these architectures. In from two different services can be contradictory. The challenge
order to standardize communications in the IoT, the OneM2M will be to manage multiple simultaneous object access. The
consortium proposed in 2015 a Restful architecture based ultimate challenge is to define formally a model transformation
on resources to facilitate interoperability between objects [1] approach that allows an automated development of services for
[16]. Guinard & al. proposed a resource based approach to smart homes.
help break the technological silos of communication between
R EFERENCES
connected objects in the IoT [13] [12]. Our architectural
choices are consistent with those have been made in [16], [1] Onem2m. http://www.onem2m.org/. Accessed: 2016-04-30.
[2] Rosa Alarcon and Erik Wilde. From restful services to rdf: connecting
[13] and [12]. In the first level of our architecture we choose the web and the semantic web. arXiv preprint arXiv:1006.2718, 2010.
to abstract the real world objects in a resource model to [3] Rosa Alarcon, Erik Wilde, and Jesus Bellido. Hypermedia-driven restful
standardize the wide variety of these objects. These choices service composition. In Service-Oriented Computing, pages 111–120.
Springer, 2010.
help break the communication silos since all objects, whatever [4] Tony Andrews, Francisco Curbera, Hitesh Dholakia, Yaron Goland,
their nature can communicate in the same way using the Johannes Klein, Frank Leymann, Kevin Liu, Dieter Roller, Doug Smith,
uniform interface. Satish Thatte, et al. Business process execution language for web
services, 2003.
Other work was conducted around the composition of [5] Luigi Atzori, Antonio Iera, and Giacomo Morabito. The internet of
Restful web services. An approach of adapting the Business things: A survey. Computer networks, 54(15):2787–2805, 2010.
Process Execution Language (BPEL) [4] to Restfuls services [6] Rayhana Baghli and Bruno Traverson. Verbalization of business rules:
Application to ocl constraints in the utility domain. In Model-Driven
was proposed in [3] and [14]. Also, [18] propose a Topology Engineering and Software Development (MODELSWARD), 2014 2nd
and Orchestration Specification for Cloud Applications. International Conference on, pages 348–355. IEEE, 2014.
Other work focuses on semantics in IoT, [2] proposes to [7] Kamal Bhattacharya, Cagdas Gerede, Richard Hull, Rong Liu, and
Jianwen Su. Towards formal analysis of artifact-centric business process
increase resources with semantic aspects in order to make them models. In Business Process Management, pages 288–304. Springer,
accessible to machines and humans. 2007.
Our approach tends to group interesting ideas that have been [8] Kamal Bhattacharya, Richard Hull, Jianwen Su, et al. A data-centric
design methodology for business processes. Handbook of Research on
proposed by these teams in the same architecture, adapt and Business Process Modeling, pages 503–531, 2009.
complement them in order to break the current technology [9] Christian Bizer, Tom Heath, and Tim Berners-Lee. Linked data-the
silos of IoT. The architecture we propose is a horizontal trans- story so far. Semantic Services, Interoperability and Web Applications:
Emerging Concepts, pages 205–227, 2009.
verse architecture that describes how services could interact [10] Thomas Erl. Service-oriented architecture: concepts, technology, and
with connected objects and with end users. design. Pearson Education India, 2005.
[11] Roy Fielding. Fielding dissertation: Chapter 5: Representational state
V. C ONCLUSION AND FUTURE WORK transfer (rest). Recuperado el, 8, 2000.
[12] Dominique Guinard, Vlad Trifa, Friedemann Mattern, and Erik Wilde.
In this paper, we introduced a context aware architec- From the internet of things to the web of things: Resource-oriented
ture which is composed of three levels of abstraction, each architecture and best practices. In Architecting the Internet of Things,
pages 97–129. Springer, 2011.
level helps to break the current technology silos in IoT. [13] Dominique Guinard, Vlad Trifa, and Erik Wilde. A resource oriented
The resources level allows to represent the heterogeneous architecture for the web of things. In Internet of Things (IOT), 2010,
objects of IoT in a standardized form that is the resource. pages 1–8. IEEE, 2010.
[14] Cesare Pautasso. Bpel for rest. In Business Process Management, pages
This representation allows to homogenize the wide variety 278–293. Springer, 2008.
of connected objects and interactions between resources are [15] Leonard Richardson and Sam Ruby. RESTful web services. O’Reilly
done using a uniform interface. The artifacts level abstracts Media, Inc., 2008.
[16] Jorg Swetina, Guang Lu, Patricia Jacobs, Francois Ennesser, and JAESE-
resources and offers context aware views to services. This UNG Song. Toward a standardized common m2m service layer platform:
level helps to break the silo at the composition level since it Introduction to onem2m. Wireless Communications, IEEE, 21(3):20–26,
allows multiple services to access the same objects and share 2014.
[17] SBVR Team et al. Semantics of business vocabulary and rules (sbvr).
data they produce. The semantic level helps use standardized Technical report, Technical Report dtc/06–03–02, Object Management
business vocabularies and brings all the semantic aspects that Group, Needham, Massachusetts, 2006.
are necessary in the IoT. Business vocabularies are shared and [18] OASIS TOSCA. Topology and orchestration specification for cloud
applications (tosca) primer version 1.0, 2013.
understandable by non-expert users. So, the semantic level
facilitates user interactions with services.
One perspective to our work is to apply the proposed archi-
tecture to real connected objects. It will permit to check the
behavior of these objects in an environment close to real world.
These experiments will permit to to measure the performance
criteria related to the proposed architecture. A second per-
spective is to place this architecture in a service composition
process where multiple services could share several objects
simultaneously. In that context, other type of issues could
raise such as simultaneous access and modification requests
of resources. In this case, two modification requests that come

187