Classes and ML-style modules are complementary. The former are better at extension, the latter at abstraction. We propose an object-oriented calculus based on a nominal module system with mutual recursion. Our calculus is kept simple by using the same construct for modules, functors, signatures and classes. Static guarantee of safety properties turns out to be a non trivial problem in this system, as cyclic dependencies might exist. We define a decidable type system, which ensures that module definitions are well-founded, and a call-by-value semantics, for which type soundness is proved.

View full abstract

We show that there exist translations between polymorphic lambda-calculus and calculus with existential types, which form a Galois connection and moreover a Galois embedding. From a programming point of view, we established that polymorphic functions can be interpreted by abstract data types and vice versa via proof duality.

View full abstract

Dual Calculus is a term calculus system which corresponds to classical sequent calculus LK, proposed by Philip Wadler. This system aims to clarify the calculation-meaning of cut-elimination procedure of LK, especially research of relevance with call-by-name and call-by-value strategy.
LKT is also classical sequent calculus system which restricted the symmetry of LK so that it could be regarded as a natural calculation system.
In this talk, we propose the system DCT which is corresponding to LKT obtained using the idea of Dual Calculus.

View full abstract

We show the taxonomy of techniques for "controling complexity", the major goal in software engineering. We use the knowledge in SWEBOK, priciples in software engineering, and encyclopedia of software engineering as information sources. We propose the knowlegde structure for these knowledge as an three-layered model: process, activity,and knowledge units.

View full abstract

Aspect-oriented modeling deals with cross-cutting concerns at early stages of the software development. Weaving involves abstract aspect descriptions, and needs to take into accout application specific constraints. Since the weaving is done manually, some verification is necessary between the descriptions before and after the weaving. We propose to use a role-based aspect-oriented modeling method and define a notion of weaving in a systematic way. We further discuss a use of Alloy for the verification as well as precise model descriptions.

View full abstract

We are developing an application-level scheduler for improving the performance stability of Web servers. Since Web servers are running on various platforms, its performance heavily depends on the hardware and the underlying operating system (OS). To show this fact, we measured the execution performance of heavy-loaded Web servers. We ran Tomcat, which is a well-known Web application server, on several operating systems and requested it to execute light-weight and heavy-weight servlets at the same time. Our claim is that existing OS-level schedulers cannot achieve stable performance of Web servers and thus application-level scheduler would be necessary to do that.

View full abstract

In this paper, we propose a metric for measuring the complexity of remote component-based systems by using abstraction technique. As a result of evaluation experiments with a number of EJB applications, it is found that our metric can effectively measure the complexity that reflects the system's maintainability.

View full abstract

In this paper, I propose the technique of generating the system design specification for code generation from timed-automaton which modeled real-time program and from the functional specifications.In the software development which requires real-time property, there is a problem for the pressure of the development process by demand such as time restrictions, being changed frequently. Here, if it is possible to generate a code automatically from the design specificaton, we can expect to reduce the work load in a software development process. By using this system design specification for code generation, I aim to generate automatically the skeleton code for RTSJ (Real-Time Specification for Java) which is the real-time specification extension language of the object-oriented language Java.

View full abstract

We propose a method for composition of web application components based on extended finite automata.We model the behavior of a web application with dynamic contents as concurrent extended finite automata based on state transitions invoked by the requested URLs. The extended finite automata includes variable constraint. We view a web application as a data-independent system, where variables appearing in link parameters and form inputs are attached each page.We check whether a web automaton satisfies the variable constraint exhaustively with SPIN model checking tool. Finally, we show an example of application of web automata into Struts/Tiles framework.

View full abstract

Web server sends layouts and data of Web page to display whenever requested by Web browser for the traditional Web applications; but this may drop the user experience of Web page. We propose to create the Single Page Web Application (SPA) from Web application with new JSP custom tags for SPA. Because this Web application is able to be developed by the same way as the traditional Web application, a programmer is able to develop SPA, which has the improved user experience, without any additional skill.

View full abstract

The Software development getting more and more difficult, but demand for software has been rising.Moreover, the Software is more complicated because the system has become huge.Software development must be more effective to manage according to the demands.Implementing program codes and writing documents tied together will solve those.This paper suggest how to generated program code automatically from the state transition diagram (using ObCL).In this study, the code for execution of the operation is implemented into 'State' of state transition diagram and transition of the operation is implemented into 'Transition' of state transition diagram.This study tries the increase in efficiency of the development in generating logic about the transition of the program semiautomatically.

View full abstract

We present a DNA sequence analyzing system for a cellular slime mold, Dictyostelium discoideum. The upstream sequences may include cis-elements which are involved in temporal and spatial regulation of transcription. Our goal is to identify the cis-elements with statistic methods. For this purpose, we have developed a distributed system. The system consists of an alignment program based on dynamic programming which estimates candidates of cis-elements on upstream sequences, and a statistical analyzer which checks if the candidate elements have a statistical property. The system supports SOAP and the components can be deployed on the web. In this paper, we mainly discuss the system design and its efficiency.

View full abstract

When we sort a large data base where a key is a long string of characters, reducing the number of key comparisons is very important for a sorting method. In this paper, we show that the number of key comparisons of LOAS (Leaves Optimal Adaptive Sort) is much smaller than that of MERGESORT and QUICKSORT when the number of leaves of a given data is small. Here, the leaf of a sequence stands for an element in the sequence that does not have any smaller neighbor than it. Since LOAS is a merge-based sorting algorithm, it is very important to use an efficient merging algorithm. We propose LLCP merging method which drastically reduces the number of key accesses when a key is a long character string.

View full abstract

Radix sort is well known to be theoretically efficient in sorting string keys. But Multikey Quicksort (MKQ) proposed by Bentley and Sedgewick and Burstsort run faster than radix sort was reported. The main reason why radix sort defeated is that it causes much more cache misses than the others during its passes. In this paper, we propose CRadix sort, which can reduce the number of cache misses during key access. In the evaluation of 30,000,000 randomly generated string data with other 4 string sorting algorithms, Burstsort, MKQ, CMKQ (a cache optimized version of MKQ by our technique) and MSD Radix sort (MSDR), CRadix sort outperforms Burstsort , MSDR, and MKQ approximately by more than 200%, 250%, and 600% respectively.

View full abstract

There is need to design a neighborhood functionto apply local search to a combinatorial optimization problem.In this paper,we propose a method to automatically generatea neighborhood function from the description of a given problem.The method uses logical ORs in the constraints of a given problem andan existing constraint programming system as its subsystem.With the method, one can apply local search toany combinatorial optimization problems which contain ORswithout programming neighborhood functions.Experimental results showed thatlocal search solvers using neighborhood functionsgenerated automatically by the method canfind better solutions more quickly thanglobal search solvers such as clp(FD).

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

We present an embedded domain specific language for specifyingprogram transformations. The language is implemented as a monadic combinatorlibrary in Haskell. The transformations are done at compile time using themechanism of Template Haskell. The library provides a modular way to structureabstract and intuitive transformation strategies by higher-order matching andmonadic programming.

View full abstract

The notion of \Sigma-monoids is proposed by Fiore, Plotkin and Turi, to give abstract algebraic model of languages with variable binding and substitutions. In this paper, we give a free construction of \Sigma-monoids. The free \Sigma-monoid over a given presheaf serves a well-structured term language involving binding and substitutions. Moreover, the free \Sigma-monoid naturally contains interesting syntactic objects which can be viewed as "metavariables" and "environments". We analyse the term language of the free \Sig-monod by relating it with several concrete systems, especially the \lmd-calculus extended with contexts.

View full abstract

This paper shows that program fusion transformation can be easilyobtained through the rewriting induction. Since the rewritinginduction is a powerful automated deduction technique for provinginductive theorems, the method presented offers a unified framework in which a program transformation system and a automated deduction system aretightly combined. Our implementation based on S-expressionrewriting systems can treat fusion transformation for higher-ordersystems.

View full abstract

Points-to analysis is a fundamental analysis technique whose results are useful in compiler optimization and software engineering tools. Although many points-to analysis algorithms have been proposed for procedural and object-oriented languages such as C and Java, there is no points-to analysis for aspect-oriented languages until now. In this paper, we propose a flow- and context-insensitive point-to analysis for AspectJ, which is based on an Andersen-style points-to analysis for Java. The key idea of our analysis is to perform the points-to analysis crossing the boundry of aspects and classes, and therefore our analysis can handle the specific issues ralted to asepct-oriented programs.

View full abstract

We propose an aspect-oriented language for a typed functional language. The language design is based on AspectJ, the aspect-oriented language for Java, and has two features, pointcut-advice and type-extension. In this paper, we describe the design of these features and discuss the specific problems and solutions for functional languages.

View full abstract

Design by Contract is a technique to design correct softwares by specifying constraints as assertions that hold before and after a software component executes.By using this approach, the more assertion we added, the more complicated the system will be.We found that it is hard to preserve consistency on this way.To solve this problem, we introduce Aspect-Oriented modularization method for specification language.This is the method which split a large specification to some partial parts, which we take as aspect.Then we propose the translation method of the partial parts, aspect, to the State-transition model.

View full abstract

We have developed Drms and Drv, two prototypes of software failure analyzer. Drms detects, as possible sources of failure, anomalous sequences of method invocations, while Drv detects anomalous values in variables by dynamically analyzing the behavior of Java programs. Callers of anomalous method invocations and (anomalous) input values causing the anomalous invocations can then be thought of as root causes of the failure.
This paper illustrates the result of applying two prototypes to sample software failure, and discusses the possibility of applying our framework to software running in reality. Relationship between anomaly and failure is also discussed.

View full abstract

Reverse execution of programs is a function that allows programs to execute backward in time. A reversible debugger is a debugger that provides such functionality. In this paper, we propose an approach that enables reverse execution of programs written in the C language. Our approach is based on a method called dynamic translation. In dynamic translation, programs are executed while being translated on a virtual machine. Our approach has the following advantages compared with existing approaches: (1) our approach allows faster execution, (2) our approach enables reverse execution of arbitrary code as long as there are sufficient memory, and (3) our approach is very flexible. We implemented a prototype reversible debugger and evaluated its efficiency.

View full abstract

Interactive systems offer undo facilities to users for error recovery,however their scopes are confined to each application. SBUML is an extension of User-Mode Linux virtual machine that can save and restore multiple Linux computation states as snapshots. By using SBUML, it will be possible to realize a more general undo facility at the operating system level. Unlike ordinary undo facilities of applications, it enables users to undo otherwise irreversible operations such as the installation of a server. Various applications based on OS-level undo can be thought of, not only effective backup. As one of such applications, we propose an extension of the GNU debugger and present examples of how it can make debugging easier.

View full abstract

Proxy certificate is introduced to ensure dynamic delegation andsingle-sign-on for grid computing. With delegation, proxy certificate path verification is needed foreach mutual authentication process. Apparently it is time-consuming, mechanically repeated. In particular, it is hard to keep a long proxy certificate path.In this paper, we propose Proxy Certificate Trust List (PCTL) that records the delegation trace and simplifies the certificate path verification. The mechanism of PCTL is based on the traditional Public Key Infrastructure (PKI), and the required modification on SSL/TLS protocol and Lightweight Directory Access Protocol (LDAP) schema is kept minimum.

View full abstract

In this research, we proposes predator model of worms infection with protective state. Predator model includes state transition of each host and mathematical model of number of infected hosts. In previous research, it is reported that good worms will be wipe out all bad worms. But by analyzing real log of network, W32.Blaster.Worm(Bad worms) never extinct by predation of W32. Welchia.Worm(Good worms). This is not well explained by previous modeling. We proposes new model that can well explain the suviving Blaster.

View full abstract

E-Commerce (electronic commerce) has spread quickly with the spread ofIT infrastructures, such as ADSL.However, user ID and the password are used for the customerauthentication of usualy E-Commerce system , so a customer has tomanage user ID and a password for customer itself.Therefore, a customer will take a burden.The damage of an order of the fallaciousness by spoofing.In order to set the EC system on the Internet, an operating cost is high.Then, this research performs customer authentication by transmissionof E-Mail or Fax.An order EC site and a sales management system are dividedfurther, and the system which communicates data safely is proposedusing the closing space separated from the Internet by PPPcommunication.

View full abstract

[in Japanese]

View full abstract

The Feature of a new extended type operation language system

View full abstract

LMNtal is a simple language model based on graph rewriting that uses membranes to represent hierachies. LMNtal is able to represent various computation uniformly and simply using links and membrane hierachies. Distributed LMNtal system is achievable by mapping membranes to physical computers, though the method for implementing it correctly and efficiently is unknown. In this presentation, we discuss an implementation based on membrane hierachies. The distributed LMNtal runtime is implemented by extending the existing sequential implementation of LMNtal runtime written in Java. We have shown that distributed computing can be written simply using LMNtal.

View full abstract

Behavior software for real-time robots like humanoids is becoming more and morecomplex, and we have a desire to reduce development cost. Therefore we are strivingto make a more productive development environment for behavior software based on Lisp, which has benefits such as interactive processing and automatic memory management. To achieve this, we need a garbage collector suitable for robot softwarewhich is real-time, parallel, and asynchronous.In this presentation, we outline a real-time garbage collector we have implemented into EusLisp, Lisp implementation designed to develop robot software. We explain especially how to reduce suspension time and how to carry on garbage collector effectively without memory starvation.

View full abstract

LMNtal is a concurrent language model that features undirected links and membranes. The design of its operational semantics that takes the interaction between several features into account required lots of discussions and many-sided investigations. In this presentation, we introduce some latest features of LMNtal and discuss several issues we faced in designing the semantics.

View full abstract

Reachability analysis of timed automata is an important issue in the area of formal verification of real time systems. The existing approach to this problem constructs, from a timed automaton, a finite automaton called a zone graph, and then applies the model checking algorithm to this graph. In this paper, we propose an alternative approach to this problem. Our approach eliminates never fired transitions from the timed automaton using an algorithm which checks satisfiability of the fired condition in its transition. Since our approach does not construct zone graphs, which are usually very large, ours is expected to be more efficient than the existing approach.

View full abstract

In Web application development, Java server side programs spread widely because of the compatibility with Web and the aspect of an object-oriented language. However, it needs a large cost to verificate programs, since they must be allocated at a Web server to be operated. This papaer proposes a verification method for Web application, which verifies state transition of its components and sessions. We aim to support verification of Java server side programs. This method analyzes programs statically, and then constructs model and verifies it using Temporal Logic. Finally, We implemented this system and confirmed that it works correctly.

View full abstract

This research presents an automatic method for verification of programs. Our method abstracts the meanings of the operations of the program as temporal logics, and by using them we can efficiently judges whether the program fulfills specifications of the operations, for example a specification that a deadlock does not occur.Especially we focus on the concurrent programs. This procedure divides the program into several parts by codes that affects the operation of the program. And, by observing the motion of each individual part's threads, the procedure converts the program's semantics into logical formulas. The program is said to be correct if the logical formula used for the specifications can be deduced from the logical formulas obtained from our procedure.

View full abstract

Runtime verification is a method which dynamically verifies whether the system satisfies the desired properties at runtime by analyzing the event traces. This is useful to get more information during the program testing and operation. We present a runtime verification algorithm to check whether an event trace satisfies a specification given by the user. Our specification language uses a real time logic, which extends the linear temporal logic by adding a natural number time. Although these logics' model is a infinite state sequence, the event traces which is verified at runtime is finite. Hence we present a finite semantics. In our algorithm we verify dynamically by changing a data structure during the execution instead of creating a construction such as timed automaton before execution.

View full abstract

HyperPage is a document authoring program modernizing the once popular HyperCard concepts with new Object-oriented flavour.
Pages are defined as a Page class which inherits characteristics from ite base Page Class with some functions added.
Each separate page is an instance of Page Class with its contents defined.
Pages may contain objects such as frame, pane, control.
These pages are described in XML.
The basic principles of HyperPage will be given in this presentation.

View full abstract

This paper proposes transformation support of an XBRL document based on modifying taxonomy. An XBRL document includes taxonomy which defined the items of the financial information. Generally, taxonomy of an organization is made by extending fundamental taxonomy. When fundamental taxonomy is revised, it is necessary to revise the definition of taxonomy of the organization. The modification of taxonomy is so hard, because the number of the items is so large. So, we proposed the method of modifying taxonomy automatically. By comparing extended taxonomy with fundamental taxonomy, the modifying procedure of fundamental taxonomy into extended taxonomy is created, and the new taxonomy is reconstructed by using the procedure. This method allows us to transform an XBRL document easily.

View full abstract

PSD (Programmable Structured Documents) is a safe document processing framework for structured documents, such as XML. In the PSD framework, a document is processed by an external processor which is an independent process other than the document holding process. Therefore, we need an effective mechanism to pass the evaluated document between the processes. By-Need DOM provides a remote access method to an embedded DOM data on another process. By-Need DOM transfers the required part of documents, when the data becomes really needed by the evaluator, minimizing the total amount of transfered data. The part of documents is specified by XPath notation. Furthermore, By-Need DOM implements the minimal methods required by the PSD evaluator. This makes the code simple and highly portable.

View full abstract

PSD (Programmable Structured Document) is a framework in which structured documents are edited efficiently and safely by evaluating embedded expressions in themselves. The PSD processing system we are currently developing requires an external evaluator to get the DOM data of documents held in the editor. In this work, a method to prune DOM trees is proposed to improve the performance of document manipulations by avoiding unnecessary data communication between the editor and the external evaluator. Based on information about references given by the user, it generates a pruned DOM tree, eliminating unnecessary parts for the evaluation from the original DOM tree. The mechanism of tree pruning is explained and its efficiency is evaluated using examples.

View full abstract

Policy-based management shows good promise for application to automated distributed systems management. It is extremely difficult, however, to create policies for controlling the behavior of managed systems that are sufficiently accurate to ensure good reliability. There are several studies on policy conflicts to help system managers to create proper policies. In this paper, we focus on the concurrent processing issues of policies and propse a policy processing framework consisting of a static policy analysis method and a policy processing system. The method finds out all sets of policies that may possibly result in unexpected problems in managed systems. The system processes policies safely and efficiently, using the analytical results of the method.

View full abstract

GETA, Generic Engine for Transposable Association, is an engine to compute similarity among documents and words over a WAM (Word-Article Matrix). One of the essential computations in GETA is to find n documents/words of highest similarity toward a given query.
Dealing with huge WAMs requires distributed computation, and GETA has a distributed implementation under a certain probability model. However this implementation only allows flat division of WAM horizontally or vertically.
In this paper, we introduce data parallelism into this distributed computation, reformalize association computation in a hierarchical manner allowing arbitrary ways of WAM division, and show a quick parallel implementationwith a set of parallel skeletons over trees.

View full abstract

Skeletal parallel programming enables programmers to build a parallel program from ready-made components called skeletons (parallel primitives) for which efficient implementations are known to exist, making both the parallel program development and the parallelization process easier. Parallel programs in terms of skeletons are, however, not always efficient. To overcome this problem and make the skeletal parallel programming more practical, this paper proposes a new parallel skeleton library in C++. This system have an optimization mechanism which transforms successive calls of parallel skeletons into a single function call with the help of fusion transformation. This paper describes the implementation of the skeleton library and reports the effects of the optimization.

View full abstract

There are many researches on security enforcement mechanisms base on local security policies.Most of them suddenly abort the execution of program when they detect the security violations, which is not useful for end users.We introduce a mechanism that automatically reduces the application feature to fit the local security policies.Using static program analysis, it detects the modules that violates the policies, and cut them off.It also search the widget tree and automatically invalidate the widgets that is not suitable for the policy.We implemented a prototype system by modifying Eclipse.

View full abstract

This work evaluates the association aspects, which are proposed as an extension to AspectJ like aspect-oriented programming languages for supporting aspect instances associated to a group of objects. The primary application of association aspects is to modularize concerns that coordinate objects. However there have been no practical examples that confirms this. This work applies association aspects for building systems by integrating applications written in object-oriented language and compares usefulness against other programming styles. Concretely, we build an integrated development environment consisting of a text editor, a compiler and a software project management system.

View full abstract

We propose a class loader, which is called a negligent class loader(NCL), which can relax the version barrier between components for software evolution in Java. The version barrier is a mechanism that prevents an object of a version of a class from being assigned to a variable of another version of that class. In Java, if a class is loaded by different class loaders, different versions of the class are created and regarded as distinct types. By relaxing the version barrier between an old and a new components, the old application can evolve by receiving instances of the new version of a class. For accessing these instances securely, the NCL loads classes to be compatible with others.

View full abstract

In this paper we propose a framework for optimizing virtualization mechanisms by cooperation between applications and the underlying systems that provide virtual computing environments (virtual environment managers). Specifically, we propose an API invoked for optimizing the applications running on top of a virtual environment manager. Through the API, programmers tell their knowledge on the application to the virtual environment manager. Then the virtual environment manager optimizes itself by utilizing the knowledge. The API is composed of many functions. This paper describes two of them: a function for supporting speculative execution of virtualization and security checks and a function for specifying the resources that are likely to be accessed in the future.

View full abstract