Guarino and REA views on the service notion
Paul Johannesson
many respects different from ours.
Guarino says that a service cannot be defined only in terms of
changes. He gives an example of a snow removal service, which only
guarantees to keep some streets from snow. If it does not snow, no
changes will ever be made. The service just upholds some state of
affairs.
This is different from our CAiSE09 paper where services are defined
through their goals, which are economic events, i.e. changes in
resources.
Guarino also says that a service should not be defined only through
the positive effects it has on resources. For example, defining a hair
dress service only through the effects it has on someone's hair is not
sufficient; we may also want to put constraints on the input resources
used in the service, e.g. we only use schampoo with vegetarian
ingredients. So, constraints should be included in service
definitions.
We have certainly included constraints in definitions of services in
earlier papers, but in fact not in the CAiSE09 paper.
Guarino further states that a service should be seen as an obligation
to carry out some actions determined by a trigger. He does not view a
service as a resource.
Can we reconcile Guarino's views with a REA (resource) oriented view
on services? The following is an attempt to do so. We introduce a
number of service related notions, in particular service
specification, capability/service resource, service agreement, and
service delivery. The term "service" is not introduced but could of
course be used for any of the above concepts (this is just a
dictionary discussion).
A service model here: http://www.zumodrive.com/share/18zoMjM3YT
We introduce Service/Process Specification to be the (knowledge level)
description of a service (or process). It is defined in terms of the
effects a service has on resources but also a number of constraints.
The idea is here that an effect is something more general than an
economic event, it can also be the maintenance of some state of
affairs (Guarino's snow service example). Furthermore, constraints can
be included. It remains to be worked out what these constraints can
be.
A specification A is more abstract than a specification B if A and B
have the same effects but B has stronger constraints.
A thorny issue here is that now there is no strict distinction between
descriptions of services and processes. It is more of a continuum.
First, we can describe a service very abstractly stating only the
positive effects it has (nicer hair), than we add on constraints
thereby describing a more concrete service (only vegetarian
ingredients in input resources), and we can continue adding
constraints (using scissors not machine), and finally we arrive at a
description of a process.
A Capability (Service Resource) represents that an actor has the
ability to provide a service (execute a process) according to a
Service/Process Specification. A Capability is a Resource.
A Service Agreement is a relationship between a provider and a
consumer that gives the consumer access to the capability of the
provider. The relationship primarily consists of an obligation stating
that the provider should provide its capability to the consumer based
on some trigger condition. Thus, Service Agreement comes very close to
Guarino's view of services. If we view relationships as resources,
then Service Agreement is also a resource, and it is the resource that
is traded.
A Service Delivery is a process that upholds (there may be a better
word) a Service Agreement. For example, there is a hair dresser with a
hair dressing capability (Capability). He enters into an agreement
(Service Agreement) with a customer. Now he has an obligation to
provide his capability to the customer when she asks to (the trigger
in this case). When the customer does so, the hair dress session
(Service Delivery) starts, thereby upholding the Service Agreement.
Hans Weigand
> many respects different from ours.
>
> Guarino says that a service cannot be defined only in terms of
> changes. He gives an example of a snow removal service, which only
> guarantees to keep some streets from snow. If it does not snow, no
> changes will ever be made. The service just upholds some state of
> affairs.
> This is different from our CAiSE09 paper where services are defined
> through their goals, which are economic events, i.e. changes in
> resources.
Ok. Now I do not (yet?) agree with Guarino, although I think what has
written about commitments is quite useful.
As far as I can see, Guarino mixes up the service and the service
contract. It is ok to talk about a service contract, but there is no
difference between such a contract and a contract that applies to other
(REA) resources like goods.
A company that wants to be sure that its supplier delivers certain items
upon calling can build a frame contract. In this frame contract the
supplier makes a commitmemnt. The company may have to pay for that, or
the reciprocal commitment that the supplier is the preferred supplier.
Or a consumer may order a house or boat to be built, and the constructor
may commit to that. (there could be a triggering moment involved, e.g. the
temperature being above zero).
In both cases, what is exchanged in the end is a good, not
service. So the definition of Guariano is too broad in that sense.
I could argue that it is also too narrow. If my wife cleans up my desk
without me having asked it, it is certainly a service, but there is no
explicit commitment. One could object that this is outside the economic
domain. In the economic domain, there will always be a commitment, even if
the commitment coincides with the action.
In our paper, we use REA and focus on services in the business domain. His
examples are from public services. He says: what do we pay for when we
fund such services by our taxes? I would say that we tax payers pay for
the capability to be present. This is of value to us citizens, even if we
never make use of it. The "goal" (our term) of this "service" is nothing
more or less than the availaibility of rescue services or care services to
be delivered. Not to be confused with the rescue service actually given to
a person (tax payer or not) in a fire. For that person, the availability
of the fire brigade is of no value, he wants to get out! That is the
"goal" of the rescue service.
A capability being present is not typical for services, a manufacturer
also has a manufacturing capability (contracted or not).
What is a bit untypical (but not exceptional) in the case of public
services like the ones mentioned is that someone is willing (really
willing ;-) to pay for the capability being present. In normal business,
nobody pays for that. But, as the example of frame contracts show, it
might be.
Of course, it is always a matter of terminology. We could call the hair
dressing a *service delivery* and the hair dresser having opened a shop a
*service availability*. But my feeling is that using the word service for
service delivery is most close to ordinary speech.
[although Merriam Webster also has these meanings of service:
- administrative division in government (so then it is the
service provider)
- the facility supplying some public demand, e.g. bus service. (cf. above)
- and more ..]
> Guarino also says that a service should not be defined only through
> the positive effects it has on resources. For example, defining a hair
> dress service only through the effects it has on someone's hair is not
> sufficient; we may also want to put constraints on the input resources
> used in the service, e.g. we only use schampoo with vegetarian
> ingredients. So, constraints should be included in service
> definitions.
> We have certainly included constraints in definitions of services in
> earlier papers, but in fact not in the CAiSE09 paper.
Guarino has a point in so far that we did not define yet a *service
description*. The goal will be an element, but we do not need to restrict
it to that.
However, I think a service does not differ here from goods, for which we
may have second-order value attributes or so.
It might be good to define service description, although I am not sure
there is a general criterion for what should be included in there. One may
say it is a set of constraints, but so are the ZF-axioms in set theory.
So my claim is that the notions of capability and service agreement and
service delivery are not necessary in the service ontology. They are
already in REA - or if they are not, they should be! - and are inherited
as long a services are viewed as resources.
What I am not sure about is whether the goal of a service could be a
"maintain" goal. Or, in other wordings, whether a distinction should be
made between goals to be reached once, or multiple times, or even
continuously. It depends on the time ontology underlying REA I guess,
whether all these cases can be seen as "state".
Ok, now you can start to shoot if you want!
Hans
ba
I publish first just for archival purposes.
I have currently this view of a resource transfer (the purpose of this
is to get an understanding of the ontological entities involved).
There are two roles involved in one transfer - the consumer (C) and
the provider (P). A resource delivery runs through phases (n) where in
each phase something is added to the transfer.
(0) In phase 0 the actors involved are defined. An actor can be a
person, a group of persons, a representative for a group, an
organisation, etc. But after phase 0 we know who the actors are.
(1) In phase 1 a relation between the actors is defined. I think of
this relation as a 'committment'. For instance, if an actor has a sign
'Hairdresser' on his shop, I view this as a proposal or an (explicit)
committment by the hairdresser to dress some other actor's hair. The
actor may chose to accept the proposal. By accepting, a relation
between the actor's is formed. The actors now to take the roles of
Consumer or Producer in this relation. The hairdresser is committed to
be producer and the other actor is the consumer. And, adhearing to the
duality priciple of REA, the actor must commit to produce something
(typically money) and the hairdresser must commit to consume.
Note that an advertisment is not the only way to form a committment
relation. An actor may just as well ask another actor for a hair
cut. By accepting, the roles in the relation is determined. After
phase 1 we know who the actors are and what roles they have committed
to play in a resource transfer. Metaphorically, a committment is a
formation of a link on which resources can be transferred (incl. other
links).
(2) In phase 2 the economic resources to be transfered are
defined. Currently I think of this as (rights on) value objects
explicitly stated in a contract. This is of course too narrow as
pointed out by V. Allee. There are extra-contractual resources as
well. But for now, a hairdresser has a produce right on a hair cut and
Volvo has an ownership right on a car (and the car model
specification, which is another resource). Both must be assumed to
have the rights to transfer some or all their right on the resource to
another actor. Thus, after phase 2 we know what (potential) economic
resources exist.
(3) In phase 3 the transfers of resources are defined. I think of this
as formations of obligations. By accepting the terms of the contract
formed in phase 2, the hairdresser accepts that she is obliged to
perform a hair cut according to the terms of the contract, and the
customer accepts that he's obliged to pay for it. Thus, after phase 3
we know what is delivered and consumed, by whom, and according to
which terms. I'm not sure whether to view this as the point in which
the economic transfer can be said to be carried out. Sometimes I think
this is the case, and sometimes not.
(4) Finally, in phase 4 the economic resources are physically
delivered, i.e. it's a change of custody.
One problem of reconciling Guarino's notion of a service with ours is
that he considers a service an event while we see it as a
resource (which is definitely not an event). As can be gleaned from
the phases discussion above, almost everything can, in a natural way,
be thought of as (economic) resources. So I'm not yet convinced by his
arguments.
I will continue this discussion shortly ...
Hans Weigand
Birger,
I basically agree with your analysis, although I have some problem in
distinguishing phase 1 and 2. Can you commit to provide a resource without
having defined the resource? Or do you mean that in phase 2, the customer
and provider agree on the choice of resource (selection)?
Hans
Dr. Hans Weigand
Infolab, Tilburg University
email: H.We...@uvt.nl phone: +31 13 4662806 fax +31 13 4663069
ba
First: both Paul and myself are currently involved in regular semester
activities. That's why you may find the documentation of our internal
discussions a bit lacking. But this will improve.
The intuition behind phase 1 is that an actor accepts to be a consumer
xor a producer (for some goods or service). Nothing more. By one actor
accepting to be a producer and another to be a consumer a relation
between those two is formed. Phase 1 is about pairing
(i.e. a relation definition). I tentatively called this 'commitment'.
The intuition behind phase 2 is that the goods and/or services that a
consumer wants or a producer produces is defined. Before this, I
claim, this is unknown with certainty.
While it's fairly obvious that a hairdresser produces hair cuts, in
the general case it's not obvious what some producer produces. For
instance, a garage mechanic can be a producer of car repairs. But he
may very well also produce tractor repairs, repainting of houses, or
even hair cuts upon inquiry.
The car mechanic is approached by a consumer in his role as a
producer. But what is produced is to be defined, and this is what is
done in phase 2.
Of course phase 1 and 2 can collaps in some cases (like it more or
less do in the hair cut example).
Phase 3 is all about the transfer of resources (in the form of rights
distribution).
I wanted to see whether this very stylised scheme of what goes into a
resource transfer is useful. And also see whether it holds for both
goods and services. I have yet to think of the notion of co-production
and how Guarino
tackles this.
Paul Johannesson
For the moment, just some more reflections on emergency services. I am
trying my best to salvage Guarino's case :-).
"I would say that we tax payers pay for the capability to be present."
the availaibility of rescue services or care services to be
delivered."
capability is present? This seems too restricted. I do not just want
the capability to be there; it should be there for me (or for people I
care about). I want the capability to be there so I can use it when I
need it. I want someone to guarantee that the capability is there for
me when I need it. If we accept this analysis, then availability
becomes a relationship, essentially consisting of an obligation, where
someone guarantees that a capability is there to be used when needed.
And this is exactly what Guarino says! Availability is an obligation.
If an emergency service is not a relationship (obligation), what is it
then?
So, I am still a bit stuck with these emergency services and how to
analyse them. Maybe we should make a clear distinction between
emergency services and "ordinary" services, and analyse them
differently.
/Paul
ba
this Google group. It's good to have them readily available.
/birger
On Sep 23, 6:26 pm, Paul Johannesson <pauljohannesso...@gmail.com>
wrote:
Hans Weigand
> me when I need it. If we accept this analysis, then availability
> becomes a relationship, essentially consisting of an obligation, where
> someone guarantees that a capability is there to be used when needed.
> And this is exactly what Guarino says! Availability is an obligation.
> If an emergency service is not a relationship (obligation), what is it
> then?
I think that Guarino is too much focused on public emergency services.
They are one kind of services. His definition can be seen as the
description of the "goal" of these, and then indeed it is more than
availability. Typically it will be described in more detail, e.g. response
within 20 minutes.
Looking closer, the question is whether the "obligation" part is specific
(should be part of the goal description) or is something that the service
inherits from being a resource. The hair dresser also has an obligation to
cut my hair decentlty once he has entered the transaction; but this does
not mean that the goal of hairdressing is an obligation.
A difference between the two examples is that hairdressing refers to a
single event, and emergency service to something holding over a period of
time. But this is not an essential difference.
So I would like to stick to something like availability as the goal of the
emergency service, with the extra remark that once the administration has
entered a contract with the actor in question, indeed this actor
has an obligation to ensure this availability.
Availability can perhaps be worked out as something like "if event X
happens, then within time T, reaction R should be given"
ba
(ServiceOntologyWithSpec.ppt). It's an attempt to capture the
difference between a service/process and it's specification. The basis
is of course the service ontology presented in the CAiSE'09 paper.
It is also a first attempt to account for Guarinos views on things.
The way this is done is mainly through widening some interpretations
of the original concepts.
/birger
ba
DOLCE-speak) which of course makes you wonder whether there's a
special meaning of DOLCE's events. For reference purposes I quote a
passage from a document "WonderWeb Deliverable D18" by Guarino et. al
where those concepts are discussed.
2.4 Endurants and Perdurants
Classically, endurants (also called continuants) are characterized as
entities that are ‘in
time’, they are ‘wholly’ present (all their proper parts are present)
at any time of their ex-
istence. On the other hand, perdurants (also called occurrents) are
entities that ‘happen in
time’, they extend in time by accumulating different ‘temporal parts’,
so that, at any time
t at which they exist, only their temporal parts at t are present.12
For example, the book
you are holding now can be considered an endurant because (now) it is
wholly present,
while “your reading of this book” is a perdurant because, your
“reading” of the previous
section is not present now. Note that it is possible to distinguish
between ‘ordinary ob-
jects’ (like the book) and ‘events or process’ (like ‘the reading of
the book’) even when
the domain contains perdurants only. In this latter case, one relies
on properties that lie
outside spatio-temporal aspects. Indeed, one can assume that four-
dimensional entities do
not need to have different spatio-temporal locations. A person and its
life (both taken to
be 4D entities) share the same space-time region but differ on other
properties since, for
instance, color, race, beliefs and the like make sense for person
only.
Endurants and perdurants can be characterized in a different way.
Something is an
endurant if (i) it exists at more than one moment and (ii) its parts
can be determined
only relatively to something else (for instance time)[49]. In other
words, the distinction
is based on the different nature of the parthood relation: endurants
need a time-indexed
parthood, while perdurants do not. Indeed, a statement like “this
keyboard is part of my
computer” is incomplete unless you specify a particular time, while
“my youth is part of
my life” does not require such a specification.13
Footnotes:
12 Time-snapshots of perdurants (i.e., perdurants that are present
only for an instant, and which lackproper temporal parts) are a limit
case in this distinction.
13 If the domain of quantification contains both ‘objects’ and
‘events’, without reducing one kind of elements to the other, the
participation relation, stating that objects participates in events,
becomes fundamental. For example, a person may participate in a
discussion and a sword in a battle. This relation does not depend on
the characterization of objects. It is crucial also in a four
dimensionalist position where objects and events, although both 4D
entities, are kept distinct.
Note: The strange wording in the example of footnote 13 is the
original text.
ba
something else?
I was trekking in the mountains this summer. On the trail I walked
from cabin to cabin. The cabins are manned by a cabin host. For a fee
I can sleep and cook in the cabin or for a lesser fee I can camp in
the area and only use the cooking facilities. Note that I do not pay
for a meal, I only pay for the availability of cooking gear (pans,
gas, etc.) so that I can prepare my meal myself.
One can claim that providing the availability of cooking facilities is
a service. A cooking service. I may chose to use/consume the service
if I want to, at my own discretion. The camp host can on the other
hand never chose to not provide it. Since I've payed they are obliged
to to produce (or face a punishement).
In a sense this is an relation between me and (I guess) the
organisation that maintains the cabins (via its representative, the
host).
Note also that the triggering event (in Guarino's terms) for the
cooking service is me paying the host. Not me starting to cook.
Hans Weigand
> cooking service is me paying the host. Not me starting to cook.
I think this is a nice example where you pay for availability. Committed
to by the host when you pay (enter the contract). However, making the
analogy to Guarino's fire service, I would say that the triggering event
for use is when you start cooking - like when a citizen starts to use the
fire brigade service when she starts cooking (in another sense).
Paul Johannesson
Would you agree with the following statement:
Every business transaction results in a contract consisting of
commitments to provide services.
Examples:
When I have completed a transaction with a hair dresser there is a
commitment by the hair dresser to provide a hair dressing service (and
some commitments for me). Providing this service will result in me
having nicer hair.
When I have completed a transaction with a book seller there is a
commitment by the book seller to provide a book ownership transfer
service (and possibly a book delivery/transportation service).
Providing this service will result in me owning the book.
This line of reasoning makes transaction results quite uniform. In
natural language, however, the way of speaking is non-uniforn. We say
"I buy a book", but this means "I get the ownership of a book (and pay
for it)", which spelled out becomes "I get a book transfer ownership
service which when provided results in me owning the book".
Now for a simple thought experiment:
Suppose I have entered into a contract with an Internet book seller,
which states that the book seller shall transfer the ownership on the
book to me and deliver it to me. However, I never pick up the book at
my post office. Thus, I never get the ownership of the book. How
should we speak of this situation? We could say that the book seller
has honoured the contract, but he has not carried out the ownership
transfer service (this is not his fault as he tried but I did not do
my part in the process).
Well, accepting this, we would say for the Guarino snow service case,
we do not buy a snow service, but we enter into a contract, where the
snow ploughman is committed to provide services to us. (This is also
as I understand how Hans means to treat the case.) It might be the
case that the snow ploughman never provides any service to us (because
it never snows), but he can still have honoured the contract.
This seems to work out,
"buying a hair dress service" means "entering a contract where a hair
dress service is to be provided when the customer asks for it"
"buying emergency services" means "entering a contract where services
are to be provided under very special circumstances"
"buying goods" means "entering a contract where an ownership transfer
service is to be provided thereby resulting in an ownership for the
customer"
Natural language is good at shortening!
Hans Weigand
> commitments to provide services.
>
business transaction in REA ontology is defined in an atemporal way. A
business transaction is realized by a process. Part of the process is a
contract phase resulting in mutual commitments. Another part is the
fulfilment of these commitments.
>
> This seems to work out,
> "buying a hair dress service" means "entering a contract where a hair
> dress service is to be provided when the customer asks for it"
> "buying emergency services" means "entering a contract where services
> are to be provided under very special circumstances"
> "buying goods" means "entering a contract where an ownership transfer
> service is to be provided thereby resulting in an ownership for the
> customer"
In my view, the second case is fundamentally different. Why? Ask yourself
the question whether the transaction succeeds when the customer would not
get his hair service, would not get snow services as it does not snow, or
would not get the book. Or alternatively, whether the customer has to pay
for it in these cases.
In the other two cases, there is not a succesful transaction, and the
customer will not pay.
However, in the case of the snow service, there is no reason why the
government would not pay. The snow service has fulfilled its part, and so
should the government.
[note: your first statement about hair service is ambiguous. I interpret
it in the normal way, that we are talking about a hair service. An
alternative interpretation is that the hair dresser only commits to
availabilty of his service, and the customer commits to nothing - since he
may choose himself to ask or not. I am not sure whether such a contract
occurs in real life, but your statement can be interpreted this way]
Now I recognize that when in the hair service/book service the transaction
fails because of lack of cooperation of the customer, this is usually
problematic (note that there is nothing problematic with a snow service
that does not go out as long as there is no snow). Depending on the
contract rules, the problem is solved in different ways. By default, the
customer does not need to pay for the service, but he may have to pay a
fine for contract breach. The contract may also stipulate that the
customer may have to pay anyway, as the transaction is deemed to be
realized at contract state. An example is a hotel reservation for which
you have to pay whether you show up or not (unless you cancel in time). In
the case of ownership transfer, the situation may be even more subtle. Is
the book not picked up yet by the customer his property or not? And when
he has paid for it but still not picked it up? This may become a legal
issue when the logistic company throws it away after one year, or when
ownership of the book implies certain responsibilities (for a book, this
is rather artificial, but for an animal, or a house this may be a real
issue). As far as I know, in general the problem is solved by balancing
the interests of both parties. So the hotel will charge the non-show 1
night, but not all nights.
These problematic situations are interesting as they reveal something
about what is going on in a transaction, but they should not be mixed up
with succesful transactions. A book being delivered and paid is a
succesful transaction. The same with hair service. If not deliverd, there
is not a succesful transaction, even if the provider is compensated.
In this respect, the public services of Guarino are fundamentally
different. If there is no snow, or no fire, then this does not mean that
the service is not delivered.
Of course, it may be that the service is not delivered. This would happen
when the government can prove that the service is not available. For
instance, on the basis of tests or periodic checks. In that case, the
government may decide to stop payment or take back money already given.
Hans
Hans Weigand
Birger
Related to the phases of an exchange process, here a short paper from a
few years ago that explicitly considers capacity checking and reservation,
and a pragmatic web paper that puts this in semantic web context.
There are more phasing models of course, like OASIS.
ba
Hans Weigand wrote:
> Birger
>
> Related to the phases of an exchange process, here a short paper from a
> few years ago that explicitly considers capacity checking and reservation,
> and a pragmatic web paper that puts this in semantic web context.
was to try to sort out whether it makes any sense to talk about
services as events, as Guarino thinks, or as resources. Identifying
services and REA events is really unnatural.
>
> There are more phasing models of course, like OASIS.
/birger
Paul Johannesson
"transaction" is replaced with "what is agreed when the negotiation
phase has ended", it should be more according to my original
intention. So I agree with Hans's comments, which are very much to the
point, in particular that we can view the failure to uphold
availability as a contract violation, even though it never snows. As I
understand Hans now, we should try to model "availability" as a
resource (a service resource?).
Hans, we have only discussed your original bullet 3 (on service
ontology). Feel free to start discussion on bullet 4 also!
/Paul
Hans Weigand
> resource (a service resource?).
How to model it the best way in REA, I am not sure. My attempt: the snow
service isa service isa/typeof resource. The goal of the snow service is
making snow cleaning facilities available (cf cutting my hair short),
where snow cleaning facilities is another resource, or bunch of resources.
It would include the existence of these facilities (if not present
yet), the fact that they are ready-for-use as well as eager to
do so if needed. We could summarize this as "available".
The snow service is exchanged between some unit or private company and the
government in return of money, on the basis of a contract/SLA.
To complete the picture, one could model that these facilities are
resources used (consumed or used) in the snow cleaning as such, which is
also a service with the goal of making the street (resource) clean,
delivered to the public (no reciprocal value object - no
business transaction).
> Hans, we have only discussed your original bullet 3 (on service
> ontology). Feel free to start discussion on bullet 4 also!
>
In fact, I wanted to start today.
Hans Weigand
Please find enclosed a notebook document in which I have put some excerpts
from related work and indicate some items that should be worked out for an
envisioned CAISE10 paper.
The main questions for me at the moment are:
- is it new? There is relevant related work in the Semantic Web Service
domain, but that seems to stop at some point
- the validation assumes that the effects on the resources are desribed
in a declarative format. Is this a realistic assumption? The description
does not need to be exhaustive, but at least sufficient for the validation
process. Perhaps it should be limited to time/place and capacity info.
- related to the previous point: should the paper go into details of the
representations or stay at the architectural level?
ba
I understand the main point of a CAiSE10-paper to investigate problems
in the relation between a business service and an informational
service. In particular compositional aspects are intresting.
To concretise I offer a scenario:
I have a really small computer on my bike. This computer measures two
things - pulses from a magnet that occur as the wheel spins, and time.
We can phrase this in service terms: a pulse recording service and a
time keeping service. From this information I can derive a wealth of
information, e.g. average velocity as I travel to work.
At work we have a health care unit (an organisational unit. A nurse on
the premises, with labs and other stuff off premises). This unit is
responsible for maintaining a good health status among the staff.
Phrasing their work in service terms could perhaps be 'Health
maintenance service'. Just for the sake of argument we assume that the
health care unit uses information from the computer to monitor my
health.
Is it correct in this case to view the services of the bike computer
as informational services and the service of the health care unit as,
well, a business service as we have talked about it previously? I'd
like to think so.
Is one of the main problems to be addressed in the paper that a
composition of informational service may be incompatible with a
definition of a business service? For example, the business service
may be defined to measure my exercise to some high level of accuracy.
Now, if the pulse generator is not there or is of poor quality then a
true measurement of my average velocity cannot be done (an rough
estimation can be had from asking me). In other words, there may exist
a business service whose supporting/realizing informational services
are either lacking, inadequate (wrong), or not up to the job (not
performing well enough).
We can also construct an example where a business service (perhaps
composed) is incompatible with an informational service. In this case
though I think of incompatible in a different way: the business
service is incapable of using the support from the informational
service. The health care service gets all necessary information from
the computer but fails to deliver health maintenance for the staff.
Thus, a business level/informtional level failure can occur because a)
services at the informational level are not deliviering the support
needed/expected from the business service, and b) the business service
cannot use the informational service in a proper way. I understand
from Hans's paper proposal that this is a kind of discusson that
should be in the paper. Did I understand this correctly?
I definitely agree on the urgency to sort out the relations between
the business level services and the informational (web) services. As
alluded to from the above I think an architectural investigation would
be intresting, continuing where the caise09 finished.
/birger
ba
the caise10-paper. Yesterday I pointed at problems that may arise at
the business level from misconfigurations or failures of what I
understood to be informational services. But, there may be other ways
of understanding leveling problems (Dietz).
The hair dressing example:
Suppose a customer arrives at the hairdresser's shop to get a "Super
hair cut" for which he pays a premium. The hairdresser guarantees that
the customer will leave the shop completely satified or get his money
back.
A "Super hair cut" implies that a number of actions should be taken
(or services be performed): shampooing, massage, cutting, etc. and
they cannot be done in any order. If the ordering is incorrect, no
super hair cut is delivered.
On the business level a contract has been formed. Both the customer
and the hair dresser are committed to honour the contract. The hair
dresser is obliged to deliver a "Super hair cut" and the customer is
informational services (e.g. shampooing and massage) can fail on the
hair cutter's part. I guess fail would mean that the has not honoured
the part of delivering a hair cut.
1) the customer does not recognise the massage as being a massage.
This means that one of the component services failed thereby failing
the entire "Super hair cut".
2) the customer does not recognise that a massage before shampooing is
the correct ordering. "Everyone knows that it should be the other way
around!".
3) the customer is disturbed by the radio playing loud in the shop.
This makes the customer not "Completely satisfied" as the contract
said.
Thus several types of service deliveries at the infomational level
results in a failure at the business level. One can of course claim
that no failure has occured at the business level as long as the
customer does not need to pay (the contract is honoured), but I still
think it is right to talk about failures in this case.
/birger
Hans Weigand
Birger,
a short reaction on you health management and hair cut scenario. The kind
of failures that you describe is not the kind of things I was concerned
about for the caise10paper. See my post of Tuesday.
However, the relationship between business service and informational
service needs to be worked out better. Cf the remarks of Erik Proper at
the CAiSE09. The point is that if you have a BS X using an IS Y, then you
must be careful not to jump directly from X to Y when Y is offered by an
independent actor like an IT department. We should distinguish between
going from one level to another and from going from one actor (business)
to another (IT). When IT is an independent actor, it has its own business
concerns. And the Business department has its own informational concerns.
So what you typically get is that first you go from BS X to S X1, where
X1 is a supporting information service within the business, then as X1 is
about providing information, it can be supported by a service Y at the IT
department (or Birger's bike), but this Y itself also has a business level
(the contract between the IT dept and the business dept).
In this way, it is more complicated but you can account for
- that the information service provided is not equal to a business
service: there is a business service X1 with the responsibility of turning
the data provided by Y into something meaningful to the business
- that the IT dept has its own business goals - it also wants to be paid
for the service, to mention one thing.
I was thinking about adressing these problems in an extended version of
the CAiSE09 paper (item 2 of my agenda). I think that your (Birgers)
problems are more related to this rather than to the double-level
composition problem (item 4). Perhaps they can be related, but I don't see
yet.
Paul Johannesson
to rephrase some of it in REA terms:
When a web service is used in a process, it will always produce
information (which is a resource) Sometimes a web service does not
produce anything else than information, e.g. a web service for
currency conversion. Some web services, however, can also be used to
produce relationships (which are also resources, but not always
economic resources). For example, a web service for registering at
Amazon can be used to create a relationship between Amazon and a
customer. In this case, no specific commitments are created. But some
web services can be used to create contracts with concrete
commitments. For example, a web service to order books from Amazon.
Thus, a web service can be used to produce information and
relationships (which exist in the social world). A web service cannot
be used to produce effects in the physical world. However, a web
service can be used to create commitments that are about effects in
the physical world.
So, we need to represent: information, relationships, contents of
commitments.
A major issue is how to represent contents of commitments. In existing
web service descriptions, we will probably see that this is often done
only by some attributes, like arrival time, which is quite
rudimentary. REA seems only to provide a small part of the answer
here. Maybe we could be agnostic on the choice of ontology and allow
any ontology for representing contents of commitments.
/Paul
Hans Weigand
Paul,
you are completely right in the kind of concepts that we need to work
out these double or more levels. We don't need to restrict ourselves to
REA, but REA could be a good starting point. Here an excerpt from a paper
by Geerts and McCarthy www.msu.edu/~mccarth4/Alabama.doc
Commitments
At the end of the original REA paper, McCarthy called for extensions into
areas such as commitments (McCarthy, 1982, p.576), and the ontological
augmentations needed for this are displayed in Figure 4 where commitment
images for economic events are proposed. Ijiri (1975,p.130) defines a
commitment as an .agreement to execute an economic event in a well-defined
future that will result in either an increase of resources or a decrease
of resources.. Commitments are important economic phenomena, and we use
Ijiri.s term .executes. for the relation between them and the actual
economic events that follow them. We model the pair-wise connection of
requited commitments in a fashion similar to actual exchanges except we
substitute a reciprocal relationship between the two commitments where an
actual exchange has a duality relationship. Because of the importance of
reciprocal relationships, we take the additional step of reifying them at
a higher level of abstraction as economic agreements, and we differentiate
between two different types of agreements: contract and schedule, the
definition of which depends on the ultimate nature of the economic
exchange. A transfer executes a contract while a transformation executes
a schedule. For example, a sale executes a sales order which is part of a
contract, and a production job executes a production order which is part
of a schedule. Two additional relationships are needed to integrate the
commitments with the exchange description: reserves and partner. Reserves
is a special kind of stock-flow relationship that describes the scheduled
inflow and outflow of resources. A sales order results in a reservation of
the finished goods to be delivered, while a production order results in a
scheduled completion of finished goods. Finally, the partner relationship
is a special kind of participation relationship that describes the outside
agents participating in the commitments. We define the partner
relationship as a subtype of the outside relationship.
--
End of excerpt
So it seems that the concepts are already there. What is missing, as far
as I can see, is a language in which instantiated commitments can be
stated, although the OWL formalization of REA by Geert Poels may be
instrumental here.
Hans Weigand
I found the article of Paolo Traverso in which the problem is described
that I mentioned earlier. It is from 2009 and refers to an upcoming
project.
I think this is good starting point.
But what is a challenging research question starting from here?
Hans
maria
Hohfeld
We discussed whether Hohfeld's analysis of rights can be useful to
analyse value exchanges as well as services. This is in an early stage
but here is a summary of the discussion.
Hohfeld divides rights into four categories:
Duty (correlative Claim) - if someone is obliged to do something
Privilege - if someone is allowed to do something, and others may not
object
Power - capability to produce changes in the legal system towards
another subject
Immunity - the right of being kept untouched from other performing an
action
Here is an overview: http://plato.stanford.edu/entries/rights/
In a value transfer, a right on a resource is created for the
recipient. Is this right a privilege or a claim? Is this dependent on
the kind of resource in the value transfer? Is it sometimes (often) a
bundle of rights?
In a value transfer resulting in the ownership of goods, the right
seems to be a privilege - the recipient is allowed to use the goods in
various ways, but she is not obliged to anything in particular. The
same argument should hold for value transfers of information. What
about services? Should we say that I get a privilege on the service
resource? Or is it rather a claim I get? A claim that someone else has
to do something for me. In the latter case, we are somehow back to
Guarino, b.
Some examples and comments are given in this file. In fact, we started
by considering some simple as well as tricky e3value models to see how
they could be explicated. Figures attached under Files-menu.
Fig. 1.1 shows a simple e3value model. A book is exchanged for money.
Fig. 1.2 shows an explication of Fig. 1.1. The customer gets a right
on the book, an ownership right, which (primarily) means that she gets
privileges on the book (to read it, destroy it, etc.). And the
retailer gets rights (privileges) on money. Fig. 1.3 shows a model
including a voucher. The customer pays for the book, and in return she
gets a voucher, analysed as a claim to have the ownership of a book
transferred to her later. This is shown in the top value transaction.
In the bottom value transaction, the customer invokes her right, and
she gets ownership rights (privileges) on the book.
Fig. 2.1 shows a customer buying a haircut service from a barber. As
in the previous example, the barber gets ownership of money, see Fig.
2.2. The customer gets in return a claim on a haircut service
resource, meaning that she can request the barber to provide the
haircut service resource in accordance with a service description.
This is made explicit in Fig. 2.3, which adds one more value
transaction: the customer invokes the claim she received in the
previous value transaction, and in return she receives more beautiful
hair (the effect of the provision of the service). Well, for a haircut
you would not typically have this large a model, because the claim on
the haircut is normally invoked directly, so you do not place this in
a value transaction of its own.
Fig. 3.1 shows a customer buying a insurance service from an insurance
company. As in the previous example, the insurance company gets
ownership of money, see Fig. 3.2. The customer gets in return a claim
on an insurance service resource, meaning that she can request the
insurance company to provide the insurance service resource in
accordance with a service description. This is made explicit in Fig.
3.3, which adds one more value transaction: the customer invokes the
claim she received in the previous value transaction, and in return
she receives a reimbursement (the effect of the provision of the
insurance-service) in accordance with the terms of the insurance
service description. In the case of the insurance service, as opposed
to a haircut service, you will typically depict the larger model
because the claim on the insurance-service is normally not invoked
directly, so you need to place this in a value transaction of its own.
Now, this could provide some guidance on what are allowed on "the
arrows" in an e3value model. The point is that it is not sufficient to
have only transfers of value objects. First, transfers of value
objects mean that rights (privileges or claims) are created. Secondly,
we sometimes want to include the invocation of rights (claims). So, on
an arrow in an e3value model we might have:
1. the creation of right(s) concerning some value object
2. the invocation of some claim
3. the effect of invoking a claim (which may, in turn, be another
claim, and/or a privilege, and or a physical state change)
(We are using these numbers in the figures above.)
Above analysis might also, to return to main topic, provide some
guidance in defining the service concept and/or what exchange of
services means. Right now the analysis above seems to border to
Guarinos, i.e. a service as a _claim_ on a process-type like object
("...an obligation to carry out some actions determined by a
trigger...) Summary: In exchanges rights on resources are created.
When the resource is goods or information, the right is a privilege,
on services the right is a claim. So we do not equate a service with a
claim, we rather say that exchanging services (as opposed to exchaning
goods) means creation of claims instead of privileges, i.e. when I buy
a service I buy the right to request the provision of a service
resource according to its service description.
On 20 Sep, 19:17, Paul Johannesson <pauljohannesso...@gmail.com>
wrote:
maria
under the name hohfeld.ppt
>
> - Visa citerad text -
Hans Weigand
Maria,
thanks for these interesting questions.
My take on it for the moment
- an economic exchange always involves a claim somehow (the other party
has to deliver, whatever it may be, or he can be brought to court)
- privileges are related to goods exchange, as far as the latter include
transfer of ownership, and ownership consists of a number of rights
(=privileges), although information exchange is a bit special.
In the case of lending (a book), there are also certain privileges
involved, but no ownership. A service offered by an entertainment park or
museum consists of the privilege of walking there (without the customer
being able to raise any other claim on the park "doing" something - only
this privilege is to be granted and effectuated, otherwise the park does
not deliver)
- a special kind of service *contains* a conditional obligation/claim or
privilege - like the insurance. The insurance company pays money (a good)
under certain conditions. In other related cases, the claim is not about a
good but about a service, e.g. a mobile car help service. The resource
offered when the condition holds needs to be distinguished from the
overall thing ("the insurance", "the firebrigade"). I agree that this
overall thing should be classified as a service, but for me it is still a
special type.
- about the value models: you are right that the value model as such does
not distinguish between these different phases/aspects. It can be
stretched a bit in this direction. The REA ontology might be a better
way of expressing the situation (but not graphically).
Hans Weigand
One more remark: there are services, like insurance, that contain a
conditional obligation to transfer some resource (money), are there also
good exchanges that contain a conditional obligation to transfer some
service?
Yes, if you think of a good exchange that includes warranty service. One
could counter that the value object in this case is an aggregated
one. To which I could reply that that is true in virtually all cases,
still we classify the "thing" under its core function.
maria
Here is another take on services:
A service resource is a resource for which we must describe how it can
be used in a process.
That is all we can say in general about a service resource. This is
because a service resource is an abstract resource, so there is not
much we can say. For goods, on the other hand, we can describe them by
weight, volume, location, etc.
Here is a class diagram putting this into context:
For any resource, we may want to specify how it must be produced (but
we do not have to). For example, a hair dress service uses only
ecologic oils. Tomatoes are produced without pesticides. This is
captured by the association mustBeProducedAccordingTo.
For any resource, we may want to specify how it can be used (but we do
not have to). For example, a hair dress service can be used to improve
the beuaty of hair. Tomatoes can be used to produce lasagnas. This is
captured by the association canBeUsedAccordingTo.
Now, for a service resource, the association canBeUsedAccordingTo must
be filled. We have to tell how the service resource can be used. We
have indicated this by different multiplicities.
So, the difference between a Resource in general and a service
resource is quite small, just one difference in a multiplicity.
Above was meant as an analysis of the service concept, the diagram
with service resources related to service specifications and
constraints may also be used as an attempt to extend our Caise09-paper
in order to model Guarinos "Snow-ploughing"-service example. Right now
the Caise-paper does not take care of this case, i.e. we only state
that a service realizes the goals of the service, i.e. measured in
terms of changes (plus or minus) of resource-attributes. If no snow
falls, this service may very well have been "realized", however we
cannot prove this using the Caise-machinery. The service process
specification and constraints is meant to model how a service may be
used and conditions under which a service fulfills some goal.
A third use of the diagram (as already indicated) might be to try to
make it more clear what distinguishes a service resource from just any
other (old ;) recourse. To return to your comment from last Friday
Hans, "... there are services, like insurance, that contain a
tried to compare a car purchase (exchange of rights in this case the
priviliges that ownership implies) to a car loan (also an exchange of
rights where the privileges are a bit relaxed (or the opposite, I
cannot do what I want with the car and I have to return it). The first
case is almost certainly classified as an exchange of goods. The
latter one, however, seems to be naturally classified as a service,
you seldom speak of loans as good transfers, or at least not merely as
goods transfers, rather of loan-services. Yet what is transferred is
prettty similar in the case of loan vs purchase. What differs is how
we may use the transferred resources, and for how long, and the demand
that we return the transferred object in a loan. One could argue that
this can be modelled via Hohfelds privileges, which is partly true,
priviliges are transferred in loans as well as in purchases, but that
is not the whole story, we need to orchestrate other "happenings" with
respect to the loan i.e. the demand that the car/object be returned,
and when, as well. A serviced specification process (related to
economic events in the ordinary REA-fashion) can, or is at least meant
to in the diagram, specify this.
Another motivation for the diagram, not at all elaborated yet, was to
try to use REA-terminology more than before.
maria
Hans Weigand
Basically, I agree with everything in Maria's diagram, although perhaps
certain things could be expressed in a different way to convey the same
meaning.
Two remarks
1. Where is the further specification of the service to be done? In the
process or in the contract? Hruby would put obligations and terms (=
conditional obligations) in the contract. This is an alternative to
consider.
2. If the specification or part of it is done by referring to the process,
then I think that our CAiSE attempt of expressing this by means of the
goal was not too bad. However, the constraint need to be expressed that
this goal (a REA event) is part of the process. This is what Maria's
diagram expresses, albeit in another way. Secondly, we may have to detail
the meaning of goal - or distinguishing two aspects. A clock mending
service, when consumed, is related to a conversion event of the watch
(broken-> happy). But at the same time, when produced, it is related to a
conversion (decrement) event of other resources (labour, ..). Since
production and consumption of the service are simultaneous, one might say
that both aspects are important. The service is the interface between the
two. A hotel room service consumes the room (some of its capacity) and
produces something that makes the guest happy (to put it in general
terms). If this makes sense, we could see the two aspects as two goals (we
did not say that a service must have 1 goal). Or we could say that a
service has both a goal (or goals) and a source/sources (capability?).
This was expressed in a figure in the CAiSE paper, but not formalized.
Hans Weigand
Please find enclosed the n-th version of the paper - 15 to 16 pages.
I have tried to accomodate your comments. I did not include OCL (yet), the
question is whether there is space for that.
A critical point is the formality of the models.
Kind regards
Hans Weigand
I just dicovered that I sent you an old version yesterday, right?
This is the latest version.
Still working on it..
I also just found some work of O'Leary on REA and time/location. It
doesn't get very deep, but the "historical database" view is supported in
this work. As it is when we go back to McCarthy 1982.
Hans Weigand
I have spent quite some time on improving the models (adding arity's to
the relationships). I also got less satisfied with the term "booking". In
fact, the term "reserving" is more generally used and therefore more
suitable (whereas booking is ambiguous).
The only problem is that REA uses this term "reserve" as *any*
relationship between commitment and resource type/set. This is quite an
important notion, but not equivalent to reservation in the sense above.
Perhaps we better use the word "commmitted" instead of "reserved" for the
REA meaning, so that "reserve" only applies to the advance commitment,
the reservation service.