SOA: Two steps backwards in history
Still I am really sad because the object oriented knowledge so vital and flourishing during 1990’s have almost totally disappeared. I used to think that abstract domain modelling is pretty simple thing to do, but gradually I have been forced be concrete experience change my mind. Perhaps abstraction as a whole is very hard for majority of people.
As Grady Booch write in his book Object Oriented Design ( from page 15) about the limitations of human capacity to understand complex system, he describes the difference of functional decomposition and object oriented analysis and comes to the only obvious conclusion: the outcome of OO-analysis is at least one magnitude more simple that the corresponding functional decomposition, when we model behaviour. The reason is our general way of understanding reality. We comprehend our environment as collections of collaborative agents. That is our most natural way of understanding world. Our concepts (the whole natural language) i.e. words are merely abstraction of those agents. When we still deliberately add the level of abstraction and distribute the needed pieces of behaviour very evenly indeed for all the entities like events or otherwise static object we will end up creating a model with absolute minimum complexity from our perception perspective. The nature of each agent (here each domain class) will naturally limit the set of all possible action to very small number.
Functional decomposition is not supported with any structural elements at all and thus will eventually end up with completely arbitrary break down. When the steps in the total behaviour is increased the complexity of the arbitrary break down will explode!
In SOA the whole idea on simulation model has been abandoned. These people have returned 15 years back in history to times of structured programming! We (read ole dinosaurs) struggled in that swamp tens of year and at least I feel no need to return there any more !
As Grady Booch write in his book Object Oriented Design ( from page 15) about the limitations of human capacity to understand complex system, he describes the difference of functional decomposition and object oriented analysis and comes to the only obvious conclusion: the outcome of OO-analysis is at least one magnitude more simple that the corresponding functional decomposition, when we model behaviour. The reason is our general way of understanding reality. We comprehend our environment as collections of collaborative agents. That is our most natural way of understanding world. Our concepts (the whole natural language) i.e. words are merely abstraction of those agents. When we still deliberately add the level of abstraction and distribute the needed pieces of behaviour very evenly indeed for all the entities like events or otherwise static object we will end up creating a model with absolute minimum complexity from our perception perspective. The nature of each agent (here each domain class) will naturally limit the set of all possible action to very small number.
Functional decomposition is not supported with any structural elements at all and thus will eventually end up with completely arbitrary break down. When the steps in the total behaviour is increased the complexity of the arbitrary break down will explode!
In SOA the whole idea on simulation model has been abandoned. These people have returned 15 years back in history to times of structured programming! We (read ole dinosaurs) struggled in that swamp tens of year and at least I feel no need to return there any more !
posted by Jukka Tamminen at 11:58 AM
0 Comments:
Post a Comment
<< Home