Yes and no, Rob.

If only interface and contract/description matter, then anything and everything 
is the service. Would you agree with this?

I mean that in addition to interface +contract there should be preservation of 
SO principles. If I take an old monolithic app written Lotus Domino Java and 
wrap it with RMI interface, I'm supposed to get a service. Well, this service 
is single-threaded, has to be rebooted every night, exposes its internal 
errors, and so on, i.e. it dictates consumer what and how to do. Even if I put 
all these conditions into Service Description, I do not think there will be 
many consumers that wanted to set the Service Contract with such service. IMO, 
in Erl's books, service contract plays the role of service interface.

BTW, I do not believe that placing an additional interface on a system makes it 
service (even if the interface name contains word 'service').

- Michael



________________________________
From: Rob Eamon <[email protected]>
To: [email protected]
Sent: Friday, January 16, 2009 10:42:26 PM
Subject: [service-orientated-architecture] Re: Vaughan reviews Erl's Tome


--- In service-orientated- architecture@ yahoogroups. com, Michael Poulin 
<m3pou...@.. .> wrote:

> Can anybody explain me how a legacy system wrapped with "non-
> standard contracts " or even "standard contracts ", i.e. 
> interfaces, can become a Service while the legacy system itself 
> does not demonstrate any implementaion of SO principles? 

Because in SO, the interface and the service contract are all that 
matter. How the service is implemented is of no concern, as long as it 
does what the service description says that it will when invoked. 

If the legacy system provides the needed functionality defined in the 
service description, then it doesn't matter how ugly, convoluted, 
nasty, spaghetti-code, etc., that the implementation might be.

Service implementation has an architecture of its own.

-Rob

 


      

Reply via email to