Skip to content

Feed aggregator

Re: [UML Forum] Data-Centric vs. Object-Oriented

UML Forum Google Group - Tue, 04/09/2013 - 13:19
> notified when a message of interest is placed on a queue. Otherwise,
> the subscriber is just polling a queue and it is not
> publish/subscribe. The most efficient way to do that is via a lookup
> table to address the notifications. (Not to mention that when the
> publisher adds a message, the intermediary needs to find the right
Categories: Communities

Re: [UML Forum] how to rep composition with a set of entities?

UML Forum Google Group - Tue, 04/09/2013 - 13:19
Yes. You can apply an 'exclusive OR' constraint between the two
relationships. That would allow (1 or more Y) OR (1 or more Z). On phone
so can't look up ref at the moment, will look later. Google is also your
friend.

As an aside: a rel can't be 'composite (1:1)' and allow more than 1 (Y or
Z). It would have to be 1:Many in both cases.

Categories: Communities

how to rep composition with a set of entities?

UML Forum Google Group - Sun, 04/07/2013 - 20:48
Hello - Say entity X has a composite (1:1) relationship with entity Y and also with entity Z. However, entity X should be allowed to have more than 1 Y and more than 1 Z. Just not more than 1 unique combination of Y and Z.
Entity Y and Z are unrelated so there's no opportunity to combine Y and Z together into a YZ entity. So does UML have a proper way to represent this relationship?
Categories: Communities

Re: [UML Forum] Data-Centric vs. Object-Oriented

UML Forum Google Group - Sun, 04/07/2013 - 20:48
Responding to Mercer-Hursh...

In any publish/subscribe infrastructure, the subscriber needs to be
notified when a message of interest is placed on a queue. Otherwise, the
subscriber is just polling a queue and it is not publish/subscribe. The
most efficient way to do that is via a lookup table to address the

Categories: Communities

Re: [UML Forum] Data-Centric vs. Object-Oriented

UML Forum Google Group - Sun, 04/07/2013 - 20:48
> substantially less coupling than publish/subscribe infrastructures
> for interprocess communication, like we have been talking about,
> because the subscribing object within a subsystem has absolutely no
> knowledge of who will respond, even at the subsystem level. (Though
> the developer does when connecting the dots.)
Categories: Communities

Re: [UML Forum] Re: Data-Centric vs. Object-Oriented

UML Forum Google Group - Sun, 04/07/2013 - 20:48
"I am skeptical that [publish/subscribe at the object level] would work out
well,..." H.S. Lahman
Me, too. I've decided to stick to OOA and use DDS only as a transport. I'm
a little concerned about how it is going to work out, even at that level,
but I don't have a choice in the matter. I'll have to stick to the plan
Categories: Communities

Re: [UML Forum] Re: Data-Centric vs. Object-Oriented

UML Forum Google Group - Sun, 04/07/2013 - 20:48
> in your AAL example, I think one would have to design software quite
> differently. I am skeptical that would work out well, but I can't
> demonstrate that offhand.)

Just to be clear, I am a fan of pub/sub systems such as I have been
describing primarily for subsystem communication. There is also a role

Categories: Communities

Re: [UML Forum] Data-Centric vs. Object-Oriented

UML Forum Google Group - Sun, 04/07/2013 - 20:48
Responding to Mercer-Hursh...

But OO coupling not tight! It is as loose as you can get and still be
able to interact. An OO behavior makes and act of faith that the object
on the other end of the relationship path -- whoever it might be -- has
the data it needs. That act of faith is validated when the developer

Categories: Communities

Re: [UML Forum] Re: Data-Centric vs. Object-Oriented

UML Forum Google Group - Sun, 04/07/2013 - 20:48
Responding to George...

Mercer-Hursh and I have been going at it for years offline. If it was
hardcopy, vast forests would have been felled.

I think the notion of coupling is much more important at the 3GL level
than at the 4GL level where UML lives. The problem at the 3GL level is
physical coupling that is the direct result of the use of 3GL type

Categories: Communities

Re: Data-Centric vs. Object-Oriented

UML Forum Google Group - Sun, 04/07/2013 - 20:48
Methods can only be assessed with regard to purpose. Unfortunately UML has
nothing to say about that as it doesn't make a difference between modeling
business objects and system components.
While they sometimes be applied to the former, OO approaches have been
explicitly introduced to deal with the latter, and for that purpose are
Categories: Communities

Re: [UML Forum] Data-Centric vs. Object-Oriented

UML Forum Google Group - Sun, 04/07/2013 - 20:48
I carelessly mentioned CORBA only because it has explicit support for
message sending. My mistake for mislabeling it as OO technology.
Cheers,
Dan
Categories: Communities

Re: Data-Centric vs. Object-Oriented

UML Forum Google Group - Sun, 04/07/2013 - 20:48
Great debate!
I want to suggest that a concrete example of a loosely coupled system is a
system that must be extended w/o re-deployment of the original components.
I'm using "*loosely coupled*" to help make requirements distinct from the
design concept of "*loose coupling*." A loosely coupled system just means
Categories: Communities

Re: [UML Forum] how to model a system where we dont know the actors but only the permissions in UML

UML Forum Google Group - Sun, 03/31/2013 - 17:41
Responding to Fannader...

Since Jacobson invented use cases and actors, I think I will stick with
his view.

--
Life is the only flaw in an otherwise perfect nonexistence
-- Schopenhauer

Imagine how much more difficult physics would be if electrons had feelings
-- Richard Feynman

Categories: Communities

Data-Centric vs. Object-Oriented

UML Forum Google Group - Sun, 03/31/2013 - 17:41
Hi,
this isn't really a UML question but it has to do with modeling
methodology. The vendors for OMG DDS DCPS implementations promote
"data-centric" architecture as meeting the needs of a loosely coupled
distributed system better than OO architectures.
A key contrast between DCPS and OO architecture, such as CORBA, is that
Categories: Communities

the meaning of class/dataType

UML Forum Google Group - Sun, 03/31/2013 - 17:41
Hi ,

As i read through UML specification that:

class has a set of attributes/operations and
data type has a set of attributes/operations

1)with regards to attributes/operations of data type
could you please clarify their semantics ?
because i don't know how such a data type has attributes and operations?

Categories: Communities

Re: [UML Forum] how to model a system where we dont know the actors but only the permissions in UML

UML Forum Google Group - Sun, 03/31/2013 - 17:41
Actors in use cases should not be confused with the classes that may or may
not represent them. Those are two very different artifacts.
[link]
Remy.
Categories: Communities

Re: [UML Forum] how to model a system where we dont know the actors but only the permissions in UML

UML Forum Google Group - Sun, 03/31/2013 - 17:41
Responding to Fannader...

You are quibbling. Yes, Jacobson defined actors as classes that /could
/be viewed as roles. But he was very clear that the members of those
classes were concrete entities external to the software and those
individual entities interacted with the software, which is what I said.

Categories: Communities

Re: [UML Forum] Maintating a Sesson Use Case

UML Forum Google Group - Sun, 03/31/2013 - 17:41
Responding to Bertabs...

This sounds like you have already designed your system and are
back-filling with use cases to describe it. My advice: Don't Do That!

Use cases define requirements for the design. You develop the use cases
first and then design the system around them.

That's fine as long as you have a separate use case for

Categories: Communities

Re: [UML Forum] how to model a system where we dont know the actors but only the permissions in UML

UML Forum Google Group - Sun, 03/31/2013 - 17:41
Actors are not external entities but roles played by external entities.
[link]
[link]
[link]
Remy.
Categories: Communities

Re: [UML Forum] Your suggestions about a Class Diagram

UML Forum Google Group - Sun, 03/31/2013 - 17:41
Responding to agg...

I am at a disadvantage in reading your model since the only French I
remember is /Je ne comprend pas/. So I am mostly speculating here.
First, some general comments...

I assume that the classes LigneXxxx are subclasses of LigneDocument in a
generalization. If so, I think you need to get a better tool that can

Categories: Communities