Frank's Blog

Compatibility and Conformance 

Welcome back to the research session. This time I've been explaining the difference between compatibility and conformance and raise the issues that need urgent research. Both terms are related to the area of service-oriented architectures:

Compatibility is the property of two or more services working seamlessly together. Usually, one service is treated as the requester, meaning it is the starting point for investigation. Seamlessly means without any deadlocks. Hence, compatibility extends traditional investigations of soundness for process-internal structures to a set of services. I always rely on interaction soundness, since it's a simple - but yet powerful - kind of compatibility notion. In a sentence, interaction soundness proves a business process to be compatible with a set of predefined services that can be dynamically bound to the requesting process.

Conformance is the property of a service implementation to share certain invariants with its specification. As can already be concluded, a useful definition of conformance is quite complicated. Traditionally, we can argue that a service implementation and specification are conform if both have the same observable behavior, i.e. they are bisimulation equivalent. While bisimulation equivalence is a valid property, it is too strong regarding practical application. Consider for instance:



where an explicit choice has already been made at design-time. While this might be a wanted refinement, specification and implementation are not conformant regarding bisimulation equivalence (Since the implementation cannot receive an expertise).

Another common conformance notion is simulation. If the specification can simulate anything the implementation can do, both should be conformant. In fact, this solves the problem illustrated with bisimulation. It raises, however, compatibility issues. Consider for instance the following figure



where two specification are compatible (i.e. they always interact deadlock free) but both implementations (that conform to the specifications according to simulation) are not compatible anymore! Indeed, it would be enaugh for both specification to only have their first activity and still be similar to their specifications!

The examples illustrated some of the problems that occur using bisimulation and simulation as conformance notions. We're currently working hard to find a suitable solution.
[ view entry ] ( 1917 views ) permalink $star_image$star_image$star_image$star_image$star_image ( 3 / 2679 )
Chicago Pics 

Just as the temperature (and humidity) is melting us, here are some pictures from my Chicago trip.This has already taken place in December, with temperatures as low as -20C:


My hotel.


The local Apple store.


The Chicago river at night.


The public transportation above the streets.


Starbucks in the old part of the town.


The Chicago river once more.


The Chicago board of trade.


Skyline view from the Sears tower.
[ view entry ] ( 1901 views ) permalink $star_image$star_image$star_image$star_image$star_image ( 3 / 2598 )
My first diploma 

Unbelievable, but I received my first diploma in the age of fourteen! Actually, I participated in a competition from the famous 64'er magazine in 1992. I reached position eight by answering 52 out of 60 horrible complicated questions! Do you still know what NMI means? Or how to provide a ninth sprite on the screen? Wow, though questions years ago... Here's the proof:


[ view entry ] ( 1856 views ) permalink $star_image$star_image$star_image$star_image$star_image ( 3 / 2604 )
I'm going to be an author 

While I didn't updated this blog for almost half a year, now the time has come to tell you something new. First of all, I finished writing my thesis and handed it in for review. (That's of course my excuse ;-) But having finished a scientific work about BPM is not absolutely satisfying me. Since I always need more deadlines to work for, I now transfer my findings into a textbook. The thing will be called nothing less than "Frank Puhlmann's Business Process Management: Design, Specification, and Analysis". As already my name as part of the title suggests - in combination with the subtitle - I'm going to present my very own approach to BPM in this book. The book will be split into three major parts, where in the first part a graphical notation for designing business processes will be introduced. This notation closely resembles a subset of the BPMN - restricted for usability. The second part will introduce the formal specification of business processes. I will present a straightforward semantics for the graphical notation using the pi-calculus (is there still anything else out there?). Using the formal specification, the third part discusses the analysis of business processes regarding structural criteria. Since a well designed business process is of course free of structural error, the last chapter also closes the link with the first one, where the question of "good design" is discussed. As an appetizer, I can already show you the cover design. By the way, I took the picture in Chicago, IL, where I attended the fourth international conference on service oriented computing last December.


[ view entry ] ( 1378 views ) permalink $star_image$star_image$star_image$star_image$star_image ( 3 / 1324 )
What I Do 


This time something tagged "research". One condition of handing in a PhD thesis at my university is a generally understandable abstract. In the last hour I've got an inspiration on that. Nothing spectacular, but before I'm removing it from the final version for some reasons yet unknown to me (maybe my Prof knows), I'm giving you a clue on what I'm doing. Further reading is available here. So enjoy page one of my work:

---

This thesis deals with a fundamental question of each scientific discipline. The question that is asked is nothing less than the very how does it work? The scientific discipline investigated is Business Process Management (BPM). In a nutshell, BPM provides concepts and technologies for realizing digitalized enterprises. This includes, but is not limited to, capturing, analyzing, deploying, running, monitoring, and mining business processes. Business processes in all flavors. Ranging from ad-hoc arrangements of simple activities over production workflows up to distributed, interacting services. This thesis focuses on how to capture -- formally -- data, processes, and interactions that make up the core of BPM.

The theory that is applied to build the formal models is called the pi-calculus. This calculus supports a simple idea inherently found in any real BPM application, that has yet been overlooked. Imagine a person called Steve. Now imagine the formal model of Steve. Let's call it S. In existing theoretical approaches to BPM, S had fixed connections to all other objects or subjects he could interact with. In the real world, Steve doesn't. Now image Steve would like to to call another person called Mary. However, there's one problem. Steve doesn't have Mary's phone number. Instead, he has the phone number of the directory assistance. The assistance can handle him the number of Mary. And Steve then can directly talk to her. Image the formal model again. S has a fixed connection to the directory assistance, called A. A in turn has knowledge of all currently registered phone numbers. When A handles the connection to M, that's the formal model of Mary, to S, an interesting thing happens. The topology of the connections is changed. S gains knowledge of M. During the course of the action, a connection between S and M is established that has formerly not existed. That's called link passing mobility.

While link passing mobility seems evident, and furthermore builds the very foundation of the Internet, its theoretical treatment in the area of BPM has long been neglected. That is where this thesis hooks into. It provides an in-deep analysis on the formal representation of data, processes, and interactions based on accredited pattern catalogues. Algorithms for mapping graphical notations to the pi-calculus. Extensions for dynamic binding in graphical notations. Two new kinds of soundness and a new behavioral equivalence criterion. Proof theories therefore. And a prototypical tool chain for investigating the practical feasibility of the results.

---

Sounds like an advertisement you say? That's actually what it should be!
[ view entry ] ( 1573 views ) permalink $star_image$star_image$star_image$star_image$star_image ( 3.1 / 516 )

<<First <Back | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | Next> Last>>