S.T. Chanson and J. Zhu, "A Unified Approach to Protocol Test Sequence Generation," Proc. INFOCOM'93 Conf., vol. 1, pp. 106114, 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....71, 72] Ural applied the all uses [72, 61] criterion, used for testing software written in block structured programming languages, to Estelle [12, 13] specification of protocols. Miller and Paul [54] introduced a method to generate tests for both control and data for EFSM models. Chanson and Zhu [15] proposed a test generation method which considers both the control flow and data aspects of the EFSM models. Lee and Yannakakis [49] provided a method to convert a class of EFSMs, where input variables are assumed to have finite domains, into equivalent FSMs. For a restricted class of LOTOS ....

S. Chanson and J. Zhu. A Unified Approach to Protocol Test Sequence Generation. Proc. of IEEE INFOCOM 1993, pp. 1d.1.1-1d.1.9.

....[71 73] Ural applied the all uses [73,61] criterion, used for testing software written in block structured programming languages, to Estelle [13,14] specification of protocols. Miller and Paul [55] introduced a method to generate tests for both control and data for EFSM models. Chanson and Zhu [17] proposed a test generation method which considers both the control flow and data aspects of the EFSM models. Lee and Yannakakis [50] provided a method to convert a class of EFSMs, where input variables are assumed to have finite domains, into equivalent FSMs. For a restricted class of LOTOS ....

S. Chanson and J. Zhu. A Unified Approach to Protocol Test Sequence Generation. Proc. of IEEE INFOCOM 1993, pp. 1d.1.1-1d.1.9.

....the X machine approach are impractical in many cases. In particular, the assumption that the data processing functions are independently testable requires that the refinement into the implementation preserves the structure of the specification. This is often not the case. Several authors [MP92, CZ93] have proposed the application of data flow testing [RW85] to EFSMs in combination with control flow testing. Both these papers were based on the specification language Estelle, a formal description language based on the EFSM model. The papers use different algorithms for generating test ....

....transitions. The transfer function of the machine is tested in the usual way using status transitions, the W method, UIO sequences or one of their derivatives. The fault coverage of these methods is therefore equivalent to the combination of the state checking method used and all dupaths. In [CZ93] an optimisation of the UIO sequence approach is offered where checking sequences are only applied in converging states (states whose outward transitions lead to a common state) This is based on the intuition that any checking sequence for a state is also a checking sequence for all ....

[Article contains additional citation context not shown here]

Samuel T. Chanson and Jinsong Zhu. A unified approach to protocol test sequence generation. In IEEE INFOCOM'93, pages 106--114, 1993.

....the issue of test derivation from EFSM specifications constantly grows in spite of the fact that all problems easily become intractable for non trivial EFSMs. Most research concentrates on finding executable test cases that attempt to cover both the control flow and data flow aspects of an EFSM [10 21]. Symbolic execution and constraint solving appear as standard techniques for dealing with a general type of machines, though some researchers look for special EFSM classes, where the problems become somewhat easier and could be solved by other techniques. Few attempts have been reported so far ....

S. T. Chanson, J. Zhu, A unified approach to protocol test sequence generation, Proceedings of INFOCOM'93, vol.1, 1993, pp. 106-114.

....in Concurrent TTCN from a prime event structure. Section 6 gives concluding remarks. 2 RELATED WORK Previous work on test case generation from state machine models mainly focused on the model of a single FSM (see e.g. Sid90] or of a single EFSM (extended finite state machine) CA90] UY91] CZ93] HUK95] and others) Test case generation from systems of communicating FSM s requires different approaches. Several approaches for generating black box test cases from systems of communicating FSM s have been proposed. Methods with explicit test purposes ( GHN93] WL93] FJJV96] ensure ....

S.T. Chanson and J.-S. Zhu. A unified approach to protocol test sequence generation. In INFOCOM '93 [INF93], pages 106--114.

.... the test purpose or the fault model for the generated test cases as input in addition to the EFSM (e.g. GHN93, WL93] methods with implicit test purpose assume test purposes for the generated test cases implicitly and usually do not require supplementary inputs in addition to the EFSM (e.g. [Sar93, CA90, PG90, UY91, MP92, CZ93]) The methods with explicit test purposes require the test designer to choose what to test. Then the methods ensure that test cases consistent with the specification and the test purposes are generated. These methods offer much flexibility, but on the other hand, they require considerable manual ....

....transitions along the test sequence can not be satisfied with any input parameter values. In general, the executability problem is undecidable. However, on condition that variables have finite domains, constraint satisfaction techniques may be applied to ensure the executability of test sequences [CA90, CZ93]. Most of the methods base the test suite structure on the EFSM transitions. This leads to a dependence of the resulting test suite on the specification style: For the same system, very different test suites may be generated depending on whether the EFSM is specified in a more state oriented or a ....

S.T. Chanson and J.-S. Zhu. A unified approach to protocol test sequence generation. In INFOCOM '93 [INF93], pages 106--114.

....an implementation conforms to the standard. A protocol specification is typically broken into its control and data portion, where the control portion is modeled by an ordinary finite state machine. Most of the formal work on conformance testing addresses the problem of testing the control portion [ADLU, BS1, BS2, CZ, DSU, Gon, KSNM, MP1, SD, SL]. Typically, machines that arise in this way have a relatively small number of states (from one to a few dozen) but a large number of different inputs and outputs (50 to 100 or more) For example, the IEEE 802.2 Logical Level Control Protocol (LLC) ANSI1] has 14 states, 48 inputs (even ....

....gave an existence proof and did not provide efficient algorithms for determining the existence of and for constructing distinguishing sequences. In the last few years there has been a resurgence of activities on this topic motivated mainly by conformance testing of communication protocols [ADLU, CZ, DSU, Gon, KSNM, MP1, SL1, SL2]. Consequently, the problems of state identification and verification have resurfaced. The complexity of state identification and verification has been resolved recently [LY1] as follows. The preset distinguishing sequence and the UIO sequence problems are both PSPACEcomplete. Furthermore, there ....

S. T. Chanson and J. Zhu, "A unified approach to protocol test sequence generation", Proc. INFOCOM, pp. 106-114, 1993.

.... two used graphs that were copies of real internetworks [12, 22] one simulated a five node WAN [4] one used two real computers to be the entire internetwork [7] one used a LAN to simulate a WAN [13] and ten, about half of the papers, used random graphs without specifying how they were generated [1, 2, 3, 6, 8, 10, 14, 15, 16, 17]. Only five used clearly specified random graphs [5, 9, 18, 20, 21] The results of any simulation depend upon the model on which the simulation is carried out. Researchers have been assuming that realistic simulation results can be attained when the graphs used to get the results are far smaller ....

Samuel T. Chanson and Jinsong Zhu. A unified approach to protocol test sequence generation. In Proceedings of IEEE INFOCOM '93, 1993.

....of the mechanism are described in terms of communicating finite state machines. This makes the specification unambiguous and makes it possible to use the techniques developed in formal methods in protocols to do validation [H90] H91] verification [LY92] K93] conformance testing [UY91] MP92][CZ93][LSKUP93] and automatic implementation [K92] The anonymous credit card is built on top of a protocol to transfer funds between two accounts in different banks. The protocol uses intermediaries and cryptographic techniques to transfer trust without either bank knowing the identity of the other ....

Samuel T. Chanson and Jinsong Zhu, "A Unified Approach to Protocol Test Sequence Generation," Proceedings of IEEE INFOCOM '93.

....or better fault coverage than the one provided by existing methods mentioned earlier. In addition, this method provides a natural procedure for locating errors in test result analysis, and gives rise to a fault coverage metric for test sequences. 3.1. 7 Cyclic characterization sequence (CCS) In [Chanson 1993], the authors used a CCS for control flow analysis. CCS(i) for state s i is the concatenation of the characterizing sequence of s i (CS(s i ) which can be either a UIO, a DS or a W set) and the sequence that brings the machine back to si after the CS is applied. CCS(s i ) is always applied ....

....increased. Then, a DFG is constructed from the FSM, test paths and test sequences for testing both control and data flow, which cover all definition observation paths, are generated from the DFG and the CFG (control flow graph) and are combined carefully to generate an executable test sequence. In [Chanson 1993], the all definition use paths criterion is used to generate paths whose first and last transition are control or data dependent. In the next section, we will present some methods that combines both control and data flow techniques. 3.3 Control and data flow testing With EFSM, the traditional ....

[Article contains additional citation context not shown here]

Samuel T. Chanson, Jinsong Zhu, "A Unified Approach to Protocol Test Sequence Generation", in Proc. IEEE INFOCOM, San Francisco, March 1993.

....the problems which arise in this extension. Finally, we extend the concept of converging transitions to the EFSM model, which can aid in reducing the length of the test sequences. 1 Introduction Protocol conformance testing has been extensively studied using the finite state machine (FSM) model [9, 10, 11, 14, 17, 20, 24]. Research using this approach take advantage of the simplicity of the specification formulation and its structure to derive techniques for obtaining test sequences with various desirable properties. The specification does not involve complex algebraic problems, and the types of faults studied are ....

Samuel T. Chanson and Jinsong Zhu, "A Unified Approach to Protocol Test Sequence Generation, " Proceedings of IEEE INFOCOM '93.

....of test sequences. Finally, in Section 7, we close the paper by highlighting some future work. 2 Underlying Model The underlying model in our study is a deterministic FSM. It is used in modeling the control part of a protocol (other techniques have been proposed for testing the data part [16, 2]) A deterministic FSM can be represented by a quintuple M = Q; X; Y; ffi; where Q; X; Y are the internal states, input alphabet and output alphabet respectively. ffi (the next state function) is a mapping of Q Theta X 2 into Q, and (the output function) is a mapping of Q Theta X into Y . ....

S.T. Chanson and J. Zhu. A unified approach to protocol test sequence generation. In Proc. IEEE INFOCOM, San Francisco, March 1993.

No context found.

S.T. Chanson and J. Zhu, "A Unified Approach to Protocol Test Sequence Generation," Proc. INFOCOM'93 Conf., vol. 1, pp. 106114, 1993.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC