Search Results

Search found 16903 results on 677 pages for 'single responsibility'.

Page 472/677 | < Previous Page | 468 469 470 471 472 473 474 475 476 477 478 479  | Next Page >

  • O the Agony - Merging Scrum and Waterfall

    - by John K. Hines
    If there's nothing else to know about Scrum (and Agile in general), it's this: You can't force a team to adopt Agile methods.  In all cases, the team must want to change. Well, sure, you could force a team.  But it's going to be a horrible, painful process with a huge learning curve made even steeper by the lack of training and motivation on behalf of the team.  On a completely unrelated note, I've spent the past three months working on a team that was formed by merging three separate teams.  One of these teams has been adopting and using Agile practices like Scrum since 2007, the other was in continuous bug fix mode, releasing on average one new piece of software per year using semi-Waterfall methods.  In particular, one senior developer on the Waterfall team didn't see anything in Agile but overhead. Fast forward through three months of tension, passive resistance, process pushback, and you have seven people who want to change and one who explicitly doesn't.  It took two things to make Scrum happen: The team manager took a class called "Agile Software Development using Scrum". The team lead explained the point of Agile was to reduce the workload of the senior developer, with another senior developer and the manager present. It's incredible to me how a single person can strongly influence the direction of an entire team.  Let alone if Scrum comes down as some managerial decree onto a functioning team who have no idea what it is.  Pity the fool. On the bright side, I am now an expert at drawing Visio process flows.  And I have some gentle advice for any first-level managers: If you preside over a team process change, it's beneficial to start the discussion on how the team will work as early as possible.  You should have a vision for this and guide the discussion, even if decisions are weeks away.  Don't always root for the underdog.  It's been my experience that managers who see themselves as compassionate and caring spend a great deal of time understanding and advocating for the one person on the team who feels left out.  Remember that by focusing on this one person you risk alienating the rest of the team, allow tension to build, and delay the resolution of the problem. My way would have been to decree Scrum, force all of my processes on everyone else, and use the past three months ironing out the kinks.  Which takes us all the way back to point number one. Technorati tags: Scrum Scrum Process Scrum and Waterfall

    Read the article

  • Talking JavaOne with Rock Star Adam Bien

    - by Janice J. Heiss
    Among the most celebrated developers in recent years, especially in the domain of Java EE and JavaFX, is consultant Adam Bien, who, in addition to being a JavaOne Rock Star for Java EE sessions given in 2009 and 20011, is a Java Champion, the winner of Oracle Magazine’s 2011 Top Java Developer of the Year Award, and recently won a 2012 JAX Innovation Award as a top Java Ambassador. Bien will be presenting the following sessions: TUT3907 - Java EE 6/7: The Lean Parts CON3906 - Stress-Testing Java EE 6 Applications Without Stress CON3908 - Building Serious JavaFX 2 Applications CON3896 - Interactive Onstage Java EE Overengineering I spoke with Bien to get his take on Java today. He expressed excitement that the smallest companies and startups are showing increasing interest in Java EE. “This is a very good sign,” said Bien. “Only a few years ago J2EE was mostly used by larger companies -- now it becomes interesting even for one-person shows. Enterprise Java events are also extremely popular. On the Java SE side, I'm really excited about Project Nashorn.” Nashorn is an upcoming JavaScript engine, developed fully in Java by Oracle, and based on the Da Vinci Machine (JSR 292) which is expected to be available for Java 8.   Bien expressed concern about a common misconception regarding Java's mediocre productivity. “The problem is not Java,” explained Bien, “but rather systems built with ancient patterns and approaches. Sometimes it really is ‘Cargo Cult Programming.’ Java SE/EE can be incredibly productive and lean without the unnecessary and hard-to-maintain bloat. The real problems are ‘Ivory Towers’ and not Java’s lack of productivity.” Bien remarked that if there is one thing he wanted Java developers to understand it is that, "Premature optimization is the root of all evil. Or at least of some evil. Modern JVMs and application servers are hard to optimize upfront. It is far easier to write simple code and measure the results continuously. Identify the hotspots first, then optimize.” He advised Java EE developers to, “Rethink everything you know about Enterprise Java. Before you implement anything, ask the question: ‘Why?’ If there is no clear answer -- just don't do it. Most well known best practices are outdated. Focus your efforts on the domain problem and not the technology.” Looking ahead, Bien said, “I would like to see open source application servers running directly on a hypervisor. Packaging the whole runtime in a single file would significantly simplify the deployment and operations.”Check out a recent Java Magazine interview with Bien about his Java EE 6 stress monitoring tool here. Originally published on blogs.oracle.com/javaone.

    Read the article

  • Is there a better term than "smoothness" or "granularity" to describe this language feature?

    - by Chris Stevens
    One of the best things about programming is the abundance of different languages. There are general purpose languages like C++ and Java, as well as little languages like XSLT and AWK. When comparing languages, people often use things like speed, power, expressiveness, and portability as the important distinguishing features. There is one characteristic of languages I consider to be important that, so far, I haven't heard [or been able to come up with] a good term for: how well a language scales from writing tiny programs to writing huge programs. Some languages make it easy and painless to write programs that only require a few lines of code, e.g. task automation. But those languages often don't have enough power to solve large problems, e.g. GUI programming. Conversely, languages that are powerful enough for big problems often require far too much overhead for small problems. This characteristic is important because problems that look small at first frequently grow in scope in unexpected ways. If a programmer chooses a language appropriate only for small tasks, scope changes can require rewriting code from scratch in a new language. And if the programmer chooses a language with lots of overhead and friction to solve a problem that stays small, it will be harder for other people to use and understand than necessary. Rewriting code that works fine is the single most wasteful thing a programmer can do with their time, but using a bazooka to kill a mosquito instead of a flyswatter isn't good either. Here are some of the ways this characteristic presents itself. Can be used interactively - there is some environment where programmers can enter commands one by one Requires no more than one file - neither project files nor makefiles are required for running in batch mode Can easily split code across multiple files - files can refeence each other, or there is some support for modules Has good support for data structures - supports structures like arrays, lists, and especially classes Supports a wide variety of features - features like networking, serialization, XML, and database connectivity are supported by standard libraries Here's my take on how C#, Python, and shell scripting measure up. Python scores highest. Feature C# Python shell scripting --------------- --------- --------- --------------- Interactive poor strong strong One file poor strong strong Multiple files strong strong moderate Data structures strong strong poor Features strong strong strong Is there a term that captures this idea? If not, what term should I use? Here are some candidates. Scalability - already used to decribe language performance, so it's not a good idea to overload it in the context of language syntax Granularity - expresses the idea of being good just for big tasks versus being good for big and small tasks, but doesn't express anything about data structures Smoothness - expresses the idea of low friction, but doesn't express anything about strength of data structures or features Note: Some of these properties are more correctly described as belonging to a compiler or IDE than the language itself. Please consider these tools collectively as the language environment. My question is about how easy or difficult languages are to use, which depends on the environment as well as the language.

    Read the article

  • Much Ado About Nothing: Stub Objects

    - by user9154181
    The Solaris 11 link-editor (ld) contains support for a new type of object that we call a stub object. A stub object is a shared object, built entirely from mapfiles, that supplies the same linking interface as the real object, while containing no code or data. Stub objects cannot be executed — the runtime linker will kill any process that attempts to load one. However, you can link to a stub object as a dependency, allowing the stub to act as a proxy for the real version of the object. You may well wonder if there is a point to producing an object that contains nothing but linking interface. As it turns out, stub objects are very useful for building large bodies of code such as Solaris. In the last year, we've had considerable success in applying them to one of our oldest and thorniest build problems. In this discussion, I will describe how we came to invent these objects, and how we apply them to building Solaris. This posting explains where the idea for stub objects came from, and details our long and twisty journey from hallway idea to standard link-editor feature. I expect that these details are mainly of interest to those who work on Solaris and its makefiles, those who have done so in the past, and those who work with other similar bodies of code. A subsequent posting will omit the history and background details, and instead discuss how to build and use stub objects. If you are mainly interested in what stub objects are, and don't care about the underlying software war stories, I encourage you to skip ahead. The Long Road To Stubs This all started for me with an email discussion in May of 2008, regarding a change request that was filed in 2002, entitled: 4631488 lib/Makefile is too patient: .WAITs should be reduced This CR encapsulates a number of cronic issues with Solaris builds: We build Solaris with a parallel make (dmake) that tries to build as much of the code base in parallel as possible. There is a lot of code to build, and we've long made use of parallelized builds to get the job done quicker. This is even more important in today's world of massively multicore hardware. Solaris contains a large number of executables and shared objects. Executables depend on shared objects, and shared objects can depend on each other. Before you can build an object, you need to ensure that the objects it needs have been built. This implies a need for serialization, which is in direct opposition to the desire to build everying in parallel. To accurately build objects in the right order requires an accurate set of make rules defining the things that depend on each other. This sounds simple, but the reality is quite complex. In practice, having programmers explicitly specify these dependencies is a losing strategy: It's really hard to get right. It's really easy to get it wrong and never know it because things build anyway. Even if you get it right, it won't stay that way, because dependencies between objects can change over time, and make cannot help you detect such drifing. You won't know that you got it wrong until the builds break. That can be a long time after the change that triggered the breakage happened, making it hard to connect the cause and the effect. Usually this happens just before a release, when the pressure is on, its hard to think calmly, and there is no time for deep fixes. As a poor compromise, the libraries in core Solaris were built using a set of grossly incomplete hand written rules, supplemented with a number of dmake .WAIT directives used to group the libraries into sets of non-interacting groups that can be built in parallel because we think they don't depend on each other. From time to time, someone will suggest that we could analyze the built objects themselves to determine their dependencies and then generate make rules based on those relationships. This is possible, but but there are complications that limit the usefulness of that approach: To analyze an object, you have to build it first. This is a classic chicken and egg scenario. You could analyze the results of a previous build, but then you're not necessarily going to get accurate rules for the current code. It should be possible to build the code without having a built workspace available. The analysis will take time, and remember that we're constantly trying to make builds faster, not slower. By definition, such an approach will always be approximate, and therefore only incremantally more accurate than the hand written rules described above. The hand written rules are fast and cheap, while this idea is slow and complex, so we stayed with the hand written approach. Solaris was built that way, essentially forever, because these are genuinely difficult problems that had no easy answer. The makefiles were full of build races in which the right outcomes happened reliably for years until a new machine or a change in build server workload upset the accidental balance of things. After figuring out what had happened, you'd mutter "How did that ever work?", add another incomplete and soon to be inaccurate make dependency rule to the system, and move on. This was not a satisfying solution, as we tend to be perfectionists in the Solaris group, but we didn't have a better answer. It worked well enough, approximately. And so it went for years. We needed a different approach — a new idea to cut the Gordian Knot. In that discussion from May 2008, my fellow linker-alien Rod Evans had the initial spark that lead us to a game changing series of realizations: The link-editor is used to link objects together, but it only uses the ELF metadata in the object, consisting of symbol tables, ELF versioning sections, and similar data. Notably, it does not look at, or understand, the machine code that makes an object useful at runtime. If you had an object that only contained the ELF metadata for a dependency, but not the code or data, the link-editor would find it equally useful for linking, and would never know the difference. Call it a stub object. In the core Solaris OS, we require all objects to be built with a link-editor mapfile that describes all of its publically available functions and data. Could we build a stub object using the mapfile for the real object? It ought to be very fast to build stub objects, as there are no input objects to process. Unlike the real object, stub objects would not actually require any dependencies, and so, all of the stubs for the entire system could be built in parallel. When building the real objects, one could link against the stub objects instead of the real dependencies. This means that all the real objects can be built built in parallel too, without any serialization. We could replace a system that requires perfect makefile rules with a system that requires no ordering rules whatsoever. The results would be considerably more robust. We immediately realized that this idea had potential, but also that there were many details to sort out, lots of work to do, and that perhaps it wouldn't really pan out. As is often the case, it would be necessary to do the work and see how it turned out. Following that conversation, I set about trying to build a stub object. We determined that a faithful stub has to do the following: Present the same set of global symbols, with the same ELF versioning, as the real object. Functions are simple — it suffices to have a symbol of the right type, possibly, but not necessarily, referencing a null function in its text segment. Copy relocations make data more complicated to stub. The possibility of a copy relocation means that when you create a stub, the data symbols must have the actual size of the real data. Any error in this will go uncaught at link time, and will cause tragic failures at runtime that are very hard to diagnose. For reasons too obscure to go into here, involving tentative symbols, it is also important that the data reside in bss, or not, matching its placement in the real object. If the real object has more than one symbol pointing at the same data item, we call these aliased symbols. All data symbols in the stub object must exhibit the same aliasing as the real object. We imagined the stub library feature working as follows: A command line option to ld tells it to produce a stub rather than a real object. In this mode, only mapfiles are examined, and any object or shared libraries on the command line are are ignored. The extra information needed (function or data, size, and bss details) would be added to the mapfile. When building the real object instead of the stub, the extra information for building stubs would be validated against the resulting object to ensure that they match. In exploring these ideas, I immediately run headfirst into the reality of the original mapfile syntax, a subject that I would later write about as The Problem(s) With Solaris SVR4 Link-Editor Mapfiles. The idea of extending that poor language was a non-starter. Until a better mapfile syntax became available, which seemed unlikely in 2008, the solution could not involve extentions to the mapfile syntax. Instead, we cooked up the idea (hack) of augmenting mapfiles with stylized comments that would carry the necessary information. A typical definition might look like: # DATA(i386) __iob 0x3c0 # DATA(amd64,sparcv9) __iob 0xa00 # DATA(sparc) __iob 0x140 iob; A further problem then became clear: If we can't extend the mapfile syntax, then there's no good way to extend ld with an option to produce stub objects, and to validate them against the real objects. The idea of having ld read comments in a mapfile and parse them for content is an unacceptable hack. The entire point of comments is that they are strictly for the human reader, and explicitly ignored by the tool. Taking all of these speed bumps into account, I made a new plan: A perl script reads the mapfiles, generates some small C glue code to produce empty functions and data definitions, compiles and links the stub object from the generated glue code, and then deletes the generated glue code. Another perl script used after both objects have been built, to compare the real and stub objects, using data from elfdump, and validate that they present the same linking interface. By June 2008, I had written the above, and generated a stub object for libc. It was a useful prototype process to go through, and it allowed me to explore the ideas at a deep level. Ultimately though, the result was unsatisfactory as a basis for real product. There were so many issues: The use of stylized comments were fine for a prototype, but not close to professional enough for shipping product. The idea of having to document and support it was a large concern. The ideal solution for stub objects really does involve having the link-editor accept the same arguments used to build the real object, augmented with a single extra command line option. Any other solution, such as our prototype script, will require makefiles to be modified in deeper ways to support building stubs, and so, will raise barriers to converting existing code. A validation script that rederives what the linker knew when it built an object will always be at a disadvantage relative to the actual linker that did the work. A stub object should be identifyable as such. In the prototype, there was no tag or other metadata that would let you know that they weren't real objects. Being able to identify a stub object in this way means that the file command can tell you what it is, and that the runtime linker can refuse to try and run a program that loads one. At that point, we needed to apply this prototype to building Solaris. As you might imagine, the task of modifying all the makefiles in the core Solaris code base in order to do this is a massive task, and not something you'd enter into lightly. The quality of the prototype just wasn't good enough to justify that sort of time commitment, so I tabled the project, putting it on my list of long term things to think about, and moved on to other work. It would sit there for a couple of years. Semi-coincidentally, one of the projects I tacked after that was to create a new mapfile syntax for the Solaris link-editor. We had wanted to do something about the old mapfile syntax for many years. Others before me had done some paper designs, and a great deal of thought had already gone into the features it should, and should not have, but for various reasons things had never moved beyond the idea stage. When I joined Sun in late 2005, I got involved in reviewing those things and thinking about the problem. Now in 2008, fresh from relearning for the Nth time why the old mapfile syntax was a huge impediment to linker progress, it seemed like the right time to tackle the mapfile issue. Paving the way for proper stub object support was not the driving force behind that effort, but I certainly had them in mind as I moved forward. The new mapfile syntax, which we call version 2, integrated into Nevada build snv_135 in in February 2010: 6916788 ld version 2 mapfile syntax PSARC/2009/688 Human readable and extensible ld mapfile syntax In order to prove that the new mapfile syntax was adequate for general purpose use, I had also done an overhaul of the ON consolidation to convert all mapfiles to use the new syntax, and put checks in place that would ensure that no use of the old syntax would creep back in. That work went back into snv_144 in June 2010: 6916796 OSnet mapfiles should use version 2 link-editor syntax That was a big putback, modifying 517 files, adding 18 new files, and removing 110 old ones. I would have done this putback anyway, as the work was already done, and the benefits of human readable syntax are obvious. However, among the justifications listed in CR 6916796 was this We anticipate adding additional features to the new mapfile language that will be applicable to ON, and which will require all sharable object mapfiles to use the new syntax. I never explained what those additional features were, and no one asked. It was premature to say so, but this was a reference to stub objects. By that point, I had already put together a working prototype link-editor with the necessary support for stub objects. I was pleased to find that building stubs was indeed very fast. On my desktop system (Ultra 24), an amd64 stub for libc can can be built in a fraction of a second: % ptime ld -64 -z stub -o stubs/libc.so.1 -G -hlibc.so.1 \ -ztext -zdefs -Bdirect ... real 0.019708910 user 0.010101680 sys 0.008528431 In order to go from prototype to integrated link-editor feature, I knew that I would need to prove that stub objects were valuable. And to do that, I knew that I'd have to switch the Solaris ON consolidation to use stub objects and evaluate the outcome. And in order to do that experiment, ON would first need to be converted to version 2 mapfiles. Sub-mission accomplished. Normally when you design a new feature, you can devise reasonably small tests to show it works, and then deploy it incrementally, letting it prove its value as it goes. The entire point of stub objects however was to demonstrate that they could be successfully applied to an extremely large and complex code base, and specifically to solve the Solaris build issues detailed above. There was no way to finesse the matter — in order to move ahead, I would have to successfully use stub objects to build the entire ON consolidation and demonstrate their value. In software, the need to boil the ocean can often be a warning sign that things are trending in the wrong direction. Conversely, sometimes progress demands that you build something large and new all at once. A big win, or a big loss — sometimes all you can do is try it and see what happens. And so, I spent some time staring at ON makefiles trying to get a handle on how things work, and how they'd have to change. It's a big and messy world, full of complex interactions, unspecified dependencies, special cases, and knowledge of arcane makefile features... ...and so, I backed away, put it down for a few months and did other work... ...until the fall, when I felt like it was time to stop thinking and pondering (some would say stalling) and get on with it. Without stubs, the following gives a simplified high level view of how Solaris is built: An initially empty directory known as the proto, and referenced via the ROOT makefile macro is established to receive the files that make up the Solaris distribution. A top level setup rule creates the proto area, and performs operations needed to initialize the workspace so that the main build operations can be launched, such as copying needed header files into the proto area. Parallel builds are launched to build the kernel (usr/src/uts), libraries (usr/src/lib), and commands. The install makefile target builds each item and delivers a copy to the proto area. All libraries and executables link against the objects previously installed in the proto, implying the need to synchronize the order in which things are built. Subsequent passes run lint, and do packaging. Given this structure, the additions to use stub objects are: A new second proto area is established, known as the stub proto and referenced via the STUBROOT makefile macro. The stub proto has the same structure as the real proto, but is used to hold stub objects. All files in the real proto are delivered as part of the Solaris product. In contrast, the stub proto is used to build the product, and then thrown away. A new target is added to library Makefiles called stub. This rule builds the stub objects. The ld command is designed so that you can build a stub object using the same ld command line you'd use to build the real object, with the addition of a single -z stub option. This means that the makefile rules for building the stub objects are very similar to those used to build the real objects, and many existing makefile definitions can be shared between them. A new target is added to the Makefiles called stubinstall which delivers the stub objects built by the stub rule into the stub proto. These rules reuse much of existing plumbing used by the existing install rule. The setup rule runs stubinstall over the entire lib subtree as part of its initialization. All libraries and executables link against the objects in the stub proto rather than the main proto, and can therefore be built in parallel without any synchronization. There was no small way to try this that would yield meaningful results. I would have to take a leap of faith and edit approximately 1850 makefiles and 300 mapfiles first, trusting that it would all work out. Once the editing was done, I'd type make and see what happened. This took about 6 weeks to do, and there were many dark days when I'd question the entire project, or struggle to understand some of the many twisted and complex situations I'd uncover in the makefiles. I even found a couple of new issues that required changes to the new stub object related code I'd added to ld. With a substantial amount of encouragement and help from some key people in the Solaris group, I eventually got the editing done and stub objects for the entire workspace built. I found that my desktop system could build all the stub objects in the workspace in roughly a minute. This was great news, as it meant that use of the feature is effectively free — no one was likely to notice or care about the cost of building them. After another week of typing make, fixing whatever failed, and doing it again, I succeeded in getting a complete build! The next step was to remove all of the make rules and .WAIT statements dedicated to controlling the order in which libraries under usr/src/lib are built. This came together pretty quickly, and after a few more speed bumps, I had a workspace that built cleanly and looked like something you might actually be able to integrate someday. This was a significant milestone, but there was still much left to do. I turned to doing full nightly builds. Every type of build (open, closed, OpenSolaris, export, domestic) had to be tried. Each type failed in a new and unique way, requiring some thinking and rework. As things came together, I became aware of things that could have been done better, simpler, or cleaner, and those things also required some rethinking, the seeking of wisdom from others, and some rework. After another couple of weeks, it was in close to final form. My focus turned towards the end game and integration. This was a huge workspace, and needed to go back soon, before changes in the gate would made merging increasingly difficult. At this point, I knew that the stub objects had greatly simplified the makefile logic and uncovered a number of race conditions, some of which had been there for years. I assumed that the builds were faster too, so I did some builds intended to quantify the speedup in build time that resulted from this approach. It had never occurred to me that there might not be one. And so, I was very surprised to find that the wall clock build times for a stock ON workspace were essentially identical to the times for my stub library enabled version! This is why it is important to always measure, and not just to assume. One can tell from first principles, based on all those removed dependency rules in the library makefile, that the stub object version of ON gives dmake considerably more opportunities to overlap library construction. Some hypothesis were proposed, and shot down: Could we have disabled dmakes parallel feature? No, a quick check showed things being build in parallel. It was suggested that we might be I/O bound, and so, the threads would be mostly idle. That's a plausible explanation, but system stats didn't really support it. Plus, the timing between the stub and non-stub cases were just too suspiciously identical. Are our machines already handling as much parallelism as they are capable of, and unable to exploit these additional opportunities? Once again, we didn't see the evidence to back this up. Eventually, a more plausible and obvious reason emerged: We build the libraries and commands (usr/src/lib, usr/src/cmd) in parallel with the kernel (usr/src/uts). The kernel is the long leg in that race, and so, wall clock measurements of build time are essentially showing how long it takes to build uts. Although it would have been nice to post a huge speedup immediately, we can take solace in knowing that stub objects simplify the makefiles and reduce the possibility of race conditions. The next step in reducing build time should be to find ways to reduce or overlap the uts part of the builds. When that leg of the build becomes shorter, then the increased parallelism in the libs and commands will pay additional dividends. Until then, we'll just have to settle for simpler and more robust. And so, I integrated the link-editor support for creating stub objects into snv_153 (November 2010) with 6993877 ld should produce stub objects PSARC/2010/397 ELF Stub Objects followed by the work to convert the ON consolidation in snv_161 (February 2011) with 7009826 OSnet should use stub objects 4631488 lib/Makefile is too patient: .WAITs should be reduced This was a huge putback, with 2108 modified files, 8 new files, and 2 removed files. Due to the size, I was allowed a window after snv_160 closed in which to do the putback. It went pretty smoothly for something this big, a few more preexisting race conditions would be discovered and addressed over the next few weeks, and things have been quiet since then. Conclusions and Looking Forward Solaris has been built with stub objects since February. The fact that developers no longer specify the order in which libraries are built has been a big success, and we've eliminated an entire class of build error. That's not to say that there are no build races left in the ON makefiles, but we've taken a substantial bite out of the problem while generally simplifying and improving things. The introduction of a stub proto area has also opened some interesting new possibilities for other build improvements. As this article has become quite long, and as those uses do not involve stub objects, I will defer that discussion to a future article.

    Read the article

  • Use of Business Parameters in BPM12c

    - by Abhishek Mittal-Oracle
    With the release of BPM12c, a new feature to use Business Parameters is introduced through which we can define a business parameter which will behave as a global variable which can be used within BPM project. Business Administrator can be the one responsible to modify the business parameters value dynamically at run-time which may bring change in BPM process flow where it is used.This feature was a part of BPM10g product and was extensively used. In BPM11g, this feature is not present currently.Business Parameters can be defined in 2 ways:1. Using Jdev to define business parameters, and 2. Using BPM workspace to define business parameters.It is important to note that business parameters need to be mapped with a valid organisation unit defined in a BPM project. If the same is not handled, exceptions like 'BPM-70702' will be thrown by BPM Engine. This is because business parameters work along with organisation defined in a BPM project.At the same time, we can use same business parameter across different organisation units with different values. Business Parameters in BPM12c has this capability to handle multiple values with different organisation units defined in a single BPM project. This enables business to re-use same business parameters defined in a BPM project across different organisations.Business parameters can be defined using the below data types:1. int2. string 3. boolean4. double While defining an business parameter, it is mandatory to provide a default value. Below are the steps to define a business parameter in Jdev: Step 1:  Open 'Organization' and click on 'Business Parameters' tab.Step 2:  Click on '+' button.Step 3: Add business parameter name, type and provide default value(mandatory).Step 4: Click on 'OK' button.Step 5: Business parameter is defined. Below are the steps to define a business parameter in BPM workspace: Step 1: Login to BPM workspace using admin-username and password.Step 2: Click on 'Administration' on the right top side of workspace.Step 3: Click on 'Business Parameters' in the left navigation panel under 'Organization'. Step 4:  Click on '+' button.Step 5: Add business parameter name, type and provide default value(mandatory).Step 6: Click on 'OK' button.Step 7: Business parameter is defined. Note: As told earlier in the blog, it is necessary to define and map a valid organization ID with predefined variable 'organizationalUnit' under data associations in an BPM process before the business parameter is used. I have created one sample PoC demonstrating the use of Business Parameters in BPM12c and it can be found here.

    Read the article

  • Installing Forms and Reports on a development system

    - by Duncan Mills
    By popular demand I've resurrected / updated one of the old blog postings from Jan Carlin's Blog on GroundSide here. A recent (lengthy) post on the Forms forums chronicles the problems some of you have had installing F&R on a development machine. See the link in the headline of this post for the main one. When installing, here are some points to bear in mind: Download and install Weblogic Server first. http://www.oracle.com/technology/software/products/middleware/index.htmlFind the Forms and Reports (and Disco and Portal) zip files here. Download them to the desktop (or some other temporary directory of your choosing). Unzip both of the two zip files into the same new directory (maybe called 'stage') and check that you have 4 directories in the stage dir when you are finished unzipping: 'Disk1', 'Disk2', 'Disk3' and 'Disk4'. These folders are specified in the zip file structure and must be preserved for the setup executable to work. If you use WinZip and have a right click menu option that say "Extract to here", use that by right click-dragging the zip file onto the newly created directory. Don't use the "Extract to folder %HOME%\Desktop\ofm_pfrd...disk_1of2" option. That will get you into the trouble that was reported early in this thread. Free up as much memory as you can. Stop services and background processes and virus scanners and databases (you don't need a DB to install Forms) and other things lurking about on your machine. You can restart them when the install is done. Around 1.5 GB free real memory should do it. If it doesn't, free up more if you can. Don't change the swap space unless you know what you are doing. Let Windows handle it. A 1 GB machine will likely not be enough. You will likely need at least 2GB of RAM.Start the install with setup.exe from the 'Disk1' directoryChoose the Install and Configure option unless you have a good reason not to.Choose a unique instance name even if you deinstalled and removed the last install. I suggest using 'asinst_20090722_1' (today's date in ISO format with a running incremented number at the end if you install more than two times on a particular day).Unselect Portal and Discoverer and select the Builders you want.Unselect WebCacheUnselect OHS.Unselect the single sign-on option Check for any failures and choose the retry option if any occur. If that doesn't fix the problem, call Oracle Customer Support .

    Read the article

  • Renault under threat from industrial espionage, intellectual property the target

    - by Simon Thorpe
    Last year we saw news of both General Motors and Ford losing a significant amount of valuable information to competitors overseas. Within weeks of the turn of 2011 we see the European car manufacturer, Renault, also suffering. In a recent news report, French Industry Minister Eric Besson warned the country was facing "economic war" and referenced a serious case of espionage which concerns information pertaining to the development of electric cars. Renault senior vice president Christian Husson told the AFP news agency that the people concerned were in a "particularly strategic position" in the company. An investigation had uncovered a "body of evidence which shows that the actions of these three colleagues were contrary to the ethics of Renault and knowingly and deliberately placed at risk the company's assets", Mr Husson said. A source told Reuters on Wednesday the company is worried its flagship electric vehicle program, in which Renault with its partner Nissan is investing 4 billion euros ($5.3 billion), might be threatened. This casts a shadow over the estimated losses of Ford ($50 million) and General Motors ($40 million). One executive in the corporate intelligence-gathering industry, who spoke on condition of anonymity, said: "It's really difficult to say it's a case of corporate espionage ... It can be carelessness." He cited a hypothetical example of an enthusiastic employee giving away too much information about his job on an online forum. While information has always been passed and leaked, inadvertently or on purpose, the rise of the Internet and social media means corporate spies or careless employees are now more likely to be found out, he added. We are seeing more and more examples of where companies like these need to invest in technologies such as Oracle IRM to ensure such important information can be kept under control. It isn't just the recent release of information into the public domain via the Wikileaks website that is of concern, but also the increasing threats of industrial espionage in cases such as these. Information rights management doesn't totally remove the threat, but abilities to control documents no matter where they exist certainly increases the capabilities significantly. Every single time someone opens a sealed document the IRM system audits the activity. This makes identifying a potential source for a leak much easier when you have an absolute record of every person who's had access to the documents. Oracle IRM can also help with accidental or careless loss. Often people use very sensitive information all the time and forget the importance of handling it correctly. With the ability to protect the information from screen shots and prevent people copy and pasting document information into social networks and other, unsecured documents, Oracle IRM brings a totally new level of information security that would have a significant impact on reducing the risk these organizations face of losing their most valuable information.

    Read the article

  • Why You Should Attend MySQL Connect, and Register Now

    - by Bertrand Matthelié
    MySQL Connect is taking place on September 29 and 30 in San Francisco. The early bird discount enabling you to save US$ 500 is only running for a few more days, until July 13. Are you still wondering if you should sign up? Here are 10 reasons why you definitely should: Learn from other companies how they tackled similar challenges to the ones you’re facing. Find out what they learned along the way, and how you can save time, money and a lot of troubles by avoiding repeating the same mistakes and applying the best practices they’ve developed. You’ll get the chance to hear from organizations including PayPal, Verizon, Twitter, Facebook, Ticketmaster, Ning, Mozilla, CERN, Yahoo! and more! Don’t miss this unique opportunity to meet the engineers developing and supporting the MySQL products in a single location. You’ll be able to ask them all your questions, which can represent a huge time and money saver. Acquire detailed knowledge about InnoDB, the MySQL Optimizer, High Availability strategies, improving performance and scalability, enhancing security and numerous other topics. You’ll hear it straight "from the horse’s mouth" as well as from other MySQL experts in the ecosystem. Get a better understanding about Oracle’s MySQL strategy and about the MySQL roadmap, so you can better plan where to use the MySQL database and MySQL Cluster for your next web, cloud-based and other applications. Get hands-on experience about improving performance with the MySQL Performance Schema, about using MySQL Utilities, MySQL Cluster and a lot more with eight different Hands-On Labs. Express your ideas, engage into discussions and help influence the MySQL roadmap during Birds-of-a-feather sessions about replication, backup, query optimizations and other topics. Meet partners and learn about third party tools that could be useful in your architecture. Immerse yourself into the MySQL universe and hang out with MySQL experts for two days. The discussions as well as the relationships you will create can be priceless and help you execute on your next projects in a much better and faster way. Register Now to save US$500 by taking advantage of the Early bird discount running until July 13. We’ll have parallel tracks so you should consider sending a few team members to make the most of the event. Are you attending or planning to attend Oracle OpenWorld or JavaOne? You can add MySQL Connect to your registration for only US$100! Finally, it’s always a lot of fun to attend a MySQL conference. The passion and the energy are contagious…and you’ll likely get plenty of new ideas. You will find all information about the program in the MySQL Connect Content Catalog. We look forward to seeing you there! You can also read interviews with Tomas Ulin and Ronald Bradford about MySQL Connect. Sponsorship and exhibit opportunities are still available for the conference. You will find more information here.

    Read the article

  • IDC Analyst Report Touts Oracle–Accenture Strategic Initiative

    - by kristin.jellison
    Hi there, partners! Oracle Engineered Systems have been getting some love lately, and we want to share it with you! The market intelligence and advisory firm IDC recently released a report lauding Oracle and Accenture’s strategic initiative to route the performance and flexibility of Oracle Engineered Systems to clients. The report, "Oracle and Accenture Strategic Alliance Places Big Bet on Engineered Systems,” by Steve White, reflects a largely positive analysis of the relationship. White notes that the alliance is “one of the largest in the industry.” Under the relationship, Accenture has incorporated Oracle Engineered Systems—including Oracle Exadata Database Machine, Oracle Exalogic Elastic Cloud, Oracle SuperCluster, and Oracle Exalytics In-Memory Machine—into its leading datacenter transformation consulting services. Together, the two companies have also created bespoke platforms, such as the Accenture Foundation Platform for Oracle, which helps clients accelerate deployments on Oracle Fusion Middleware, running Oracle Exalogic Elastic Cloud and Oracle Exadata Database Machine. Oracle Engineered Systems deliver a single, engineered platform—including server to storage and networking. This makes it easier and cheaper for Accenture clients around the world to prepare their datacenters for managing, processing and analyzing the massive amounts of data they (rightly) anticipate seeing in the next decade. The new solutions can help reduce the effort and cost to migrate any vendor database to an Oracle Engineered Systems platform, which can lower the cost of ownership by up to 50 percent. For its part, Accenture has built a team of 300 consultants to implement and increase the flexibility and stability of client datacenters. This move further expands one of the fastest-growing full-service Oracle Enterprise solutions. Over 52,000 Accenture consultants are qualified to implement, upgrade and outsource the Oracle product suite. Accenture is a Diamond-level member of Oracle PartnerNetwork (OPN). For Oracle Partners, this update should give you at least two things to walk away with. First, this initiative is showing signs of success. As Marty Cole, group chief executive for Accenture’s Technology growth platform, put it, “We are seeing an increasing number of clients recognizing the value of consolidating their databases and taking advantage of the cost and performance benefits delivered by these solutions.” The pipeline is there—and not just for Accenture. Use this example to show your clients that investments in Oracle Engineered Systems are on the rise. Second, recognize that Oracle Engineered Systems represent one of the biggest platforms for growth that Oracle has to offer partners. As part of the agreement, Accenture is able to provide: Platform Readiness Assessments Platform Implementation App Rationalization Database Rationalization Managed Services These are all enablement opportunities you can offer customers under Oracle’s partner programs —to continue building the value of their investments, and the value of your relationship with Oracle. Take a read through the IDC report. To learn more about the partnership, see this press release. Happy selling! The OPN Communications Team

    Read the article

  • What Problems Are Better Solved By SOAP Over REST?

    In the battle for web service supremacy SOAP and REST have been battling for years. In my personal opinion this debate should have never existed. Yes, both forms can be used to create an interactive web service, but each form of a service was developed independent of each other to solve two different yet similar problems. Based my research and experience I would have to say that REST should be the preferred web service methodology and SOAP should only be used in specific situations. Note, I did not say that I was against SOAP, and in fact I actually like to use SOAP when it is needed. Criteria for using SOAP: Does the service need a guaranteed level of reliability and security? Did the provider and consumer of the service agreed on a standardized data exchange format? Does the service need data context and state management? If you answer yes to any of these questions, then you may want to consider SOAP as the format for the web service. Another way to look at the relationship between REST and SOAP is to look at the medical field.  For most things a general doctor or you family health care provider can acceptably treat most conditions from the case of a common cold to a broken bone. A general doctor more aligns with REST in my opinion because for most service requirements REST fulfills a projects needs, but what happens if you need more of an advanced examination, you would go to a specialist. A specialist would already have experience dealing with specific issues that you are experiencing giving them specific context to how best treat you going forward. SOAP acts more like a specialist doctor giving that they understand the context of an issue and can treat it based on the state of other patients they have already treated. An example of where I would use SOAP over REST in real life would be a single sign-on application. I n these cases I need to check validate a username and password for authentication and authorization of a web page request. This service would need to maintain state while it authenticated a user and while it validated access to a web page on a subsequent request. This service must process every request for access and not allow caching to ensure that every request is processed and the appropriate users are allowed to view selected web pages. References: Rozlog, M. (2010). REST and SOAP: When Should I Use Each (or Both)? Retrieved 11 20, 2011, from Infoq.com: http://www.infoq.com/articles/rest-soap-when-to-use-each

    Read the article

  • JavaOne Rock Star – Adam Bien

    - by Janice J. Heiss
    Among the most celebrated developers in recent years, especially in the domain of Java EE and JavaFX, is consultant Adam Bien, who, in addition to being a JavaOne Rock Star for Java EE sessions given in 2009 and 2011, is a Java Champion, the winner of Oracle Magazine’s 2011 Top Java Developer of the Year Award, and recently won a 2012 JAX Innovation Award as a top Java Ambassador. Bien will be presenting the following sessions: TUT3907 - Java EE 6/7: The Lean Parts CON3906 - Stress-Testing Java EE 6 Applications Without Stress CON3908 - Building Serious JavaFX 2 Applications CON3896 - Interactive Onstage Java EE Overengineering I spoke with Bien to get his take on Java today. He expressed excitement that the smallest companies and startups are showing increasing interest in Java EE. “This is a very good sign,” said Bien. “Only a few years ago J2EE was mostly used by larger companies -- now it becomes interesting even for one-person shows. Enterprise Java events are also extremely popular. On the Java SE side, I'm really excited about Project Nashorn.” Nashorn is an upcoming JavaScript engine, developed fully in Java by Oracle, and based on the Da Vinci Machine (JSR 292) which is expected to be available for Java 8.    Bien expressed concern about a common misconception regarding Java's mediocre productivity. “The problem is not Java,” explained Bien, “but rather systems built with ancient patterns and approaches. Sometimes it really is ‘Cargo Cult Programming.’ Java SE/EE can be incredibly productive and lean without the unnecessary and hard-to-maintain bloat. The real problems are ‘Ivory Towers’ and not Java’s lack of productivity.” Bien remarked that if there is one thing he wanted Java developers to understand it is that, "Premature optimization is the root of all evil. Or at least of some evil. Modern JVMs and application servers are hard to optimize upfront. It is far easier to write simple code and measure the results continuously. Identify the hotspots first, then optimize.”   He advised Java EE developers to, “Rethink everything you know about Enterprise Java. Before you implement anything, ask the question: ‘Why?’ If there is no clear answer -- just don't do it. Most well known best practices are outdated. Focus your efforts on the domain problem and not the technology.” Looking ahead, Bien remarked, “I would like to see open source application servers running directly on a hypervisor. Packaging the whole runtime in a single file would significantly simplify the deployment and operations.” Check out a recent Java Magazine interview with Bien about his Java EE 6 stress monitoring tool here.

    Read the article

  • What are the options for hosting a small Plone site? [closed]

    - by Tina Russell
    Possible Duplicate: How to find web hosting that meets my requirements? I’ve developed a portfolio website for myself using Plone 4, and I’m looking for someplace to host it. Most Plone hosting services seem to focus on large, corporate deployments, but I need something that I can afford on a very limited budget and fits a small, single-admin website. My understanding is that my basic options are thus: I can go with a hosting service that specifically provides Plone. I know of WebFaction, but what others exist? Also, I’d have two stipulations for a Plone hosting service: (a) It needs to use Plone 4, for which I’ve developed my site, and (b) it needs to allow me SSH access to a home directory (including the Plone configuration), so that I may use my custom development eggs and such. I could use a VPS hosting service. What are my options here? Again, I need something cheap and scaled to my level. I could use Amazon EC2 or a similar service (please tell me of any) and pay by the tiniest unit of data. I’m a little scared of this because I have no idea how to do a cost-benefit analysis between this and a regular VPS host. The advantage of this approach would be that I only pay for what I use, making it very scalable, but I don’t know how the overall cost would compare to any VPS host under similar circumstances. What factors enter into the cost of Amazon EC2? What can I expect to pay under either option for regular traffic for a new website? Which one is more desirable for when a rush of visitors drive up my bandwidth bill? One last note: I know Plone isn’t common for websites for individuals, but please don’t try to talk me out of it here; that’s a completely different subject. For now, assume I’m sticking with Plone for good. Also, I have seen the Plone hosting services list on Plone.org—it’s twenty pages long, and the first page was nothing but professional Plone consulting services that sometimes offer hosting for business clients. So, that wasn’t much help. Thank you!

    Read the article

  • Au revoir, Python?

    - by GuySmiley
    I'm an ex-C++ programmer who's recently discovered (and fallen head-over-heels with) Python. I've taken some time to become reasonably fluent in Python, but I've encountered some troubling realities that may lead me to drop it as my language of choice, at least for the time being. I'm writing this in the hopes that someone out there can talk me out of it by convincing me that my concerns are easily circumvented within the bounds of the python universe. I picked up python while looking for a single flexible language that will allow me to build end-to-end working systems quickly on a variety of platforms. These include: - web services - mobile apps - cross-platform client apps for PC Development speed is more of a priority at the time-being than execution speed. However, in order to improve performance over time without requiring major re-writes or architectural changes I think it's imperative to be able to interface easily with Java. That way, I can use Java to optimize specific components as the application scales, without throwing away any code. As far as I can tell, my requirement for an enterprise-capable, platform-independent, fast language with a large developer base means it would have to be Java. .NET or C++ would not cut it due to their respective limitations. Also Java is clearly de rigeur for most mobile platforms. Unfortunately, tragically, there doesn't seem to be a good way to meet all these demands. Jython seems to be what I'm looking for in principle, except that it appears to be practically dead, with no one developing, supporting, or using it to any great degree. And also Jython seems too married to the Java libraries, as you can't use many of the CPython standard libraries with it, which has a major impact on the code you end up writing. The only other option that I can see is to use JPype wrapped in marshalling classes, which may work although it seems like a pain and I wonder if it would be worth it in the long run. On the other hand, everything I'm looking for seems to be readily available by using JRuby, which seems to be much better supported. As things stand, I think this is my best option. I'm sad about this because I absolutely love everything about Python, including the syntax. The perl-like constructs in Ruby just feel like such a step backwards to me in terms of readability, but at the end of the day most of the benefits of python are available in Ruby as well. So I ask you - am I missing something here? Much of what I've said is based on what I've read, so is this summary of the current landscape accurate, or is there some magical solution to the Python-Java divide that will snuff these concerns and allow me to comfortably stay in my happy Python place?

    Read the article

  • Diagnosing the problem when Canon printer fails to print under Ubuntu

    - by MMA
    I understand that the issue of Canon printer under Linux has a number of posts. Actually one of these, was started by me. After inputs from others I have successfully printed using Canon LBP6000 from my Ubuntu machine for around a year. If it failed to print, restarting the daemon using this homemade script coaxed the printer to print. #!/bin/bash pkill -9 -x ccpd pkill -9 -x captmoncnabc /etc/init.d/ccpd start /etc/init.d/ccpd status Recently, I am not successful any more, or successful on a limited and sporadic basis. Sometimes it prints when turned on after logging in, sometimes when the driver is reinstalled. I keep on trying the random steps (call abracadabras) until I get success. Again, not always success comes. I frustrate on for hours only to get single page printed. I loose precious time on the issue of printing. I have read and read all the documents available over the Internet. However, if you please notice, none of the guides, articles, tutorials (these are too many to list here) seem to be dealing with diagnosing the problem when it fails to print. They tell you where to find the drivers, how to install them, or which script to run to make the installation process automatic. Yes, some of the articles or comments suggest a step to try, without any systematic order. But these fail to suggest a step based on symptoms, mostly. This morning, my Canon LBP6000 failed to print. After sometime, there was a message for system error, details of which was found to be something like this. When I search for this error (c3pldrv crashed with SIGSEGV in write ()), I find a number of articles including this one. None of these are actually helpful. Mostly, these are 'me too', 'tell me if you find anything'. Running captstatusui -P LBP6000 produced this, Yes, the printer is connected and actually turned on. I believe that there a number of frustrated Canon printer users out there like me. But there is not a step by step definitive guide to systematically diagnose a non-printing printer. Do you think that you can provide your diagnosing inputs so that a systematic document can be built? May be we will want the Ubuntu users to stay away from Canon printers. But as I believe, as a Linux user for more than fifteen years, such a scenario is not acceptable any more. May be this was acceptable in the infancy days of Linux, but not today. I am using Ubuntu 12.04, by the way, I prefer LTS versions.

    Read the article

  • Can you Trust Search?

    - by David Dorf
    An awful lot of referrals to e-commerce sites come from web searches. Retailers rely on search engine optimization (SEO) to correctly position their website so they can be found. Search on "blue jeans" and the results are determined by a semi-secret algorithm -- in my case Levi.com, Banana Republic, and ShopStyle show up. The NY Times recently uncovered a situation where JCPenney, via third-parties hired to help with SEO, was caught manipulating search results so they were erroneously higher in page rankings. No doubt this helped drive additional sales during this part Christmas. The article, The Dirty Little Secrets of Search, is well worth reading. My friend Ron Kleinman started an interesting discussion at the ARTS Linkedin forum. He posed the question: The ability of a single company to "punish" any retailer (by significantly impacting their on-line sales volume) who does not play by their rules ... is this a good thing or a bad thing? Clearly JCP was in the wrong and needed to be punished, but should that decision lie with Google alone? Don't get me wrong -- I'm certainly not advocating we create a Department of Search where bureaucrats think of ways to spend money, but Google wields an awful lot of power in this situation, and it makes me feel uncomfortable. Now Google is incorporating more social aspects into their search results. For example, when Google knows its me (i.e. I'm logged in when using Google) search results will be influenced by my Twitter network. In an effort to increase relevance, the blogs and re-tweeted articles from my network will be higher in the search results than they otherwise would be. So in the case of product searches, things discussed in my network will rise to the top. Continuing my blue jean example, if someone in my network had been discussing Macy's perhaps they would now be higher in the result set. soapbox: I already have lots of spammers posting bogus comments to this blog in an effort to create additional links to their sites and thus increase their search ranking. Should I expect a similar situation in Twitter and eventually Facebook? Now retailers need to expand their SEO efforts to incorporate social media as well, but do us all a favor and please don't cheat.

    Read the article

  • Slow Firefox Javascript Canvas Performance?

    - by jujumbura
    As a followup from a previous post, I have been trying to track down some slowdown I am having when drawing a scene using Javascript and the canvas element. I decided to narrow down my focus to a REALLY barebones animation that only clears the canvas and draws a single image, once per-frame. This of course runs silky smooth in Chrome, but it still stutters in Firefox. I added a simple FPS calculator, and indeed it appears that my page is typically getting an FPS in the 50's when running Firefox. This doesn't seem right to me, I must be doing something wrong here. Can anybody see anything I might be doing that is causing this drop in FPS? <!DOCTYPE HTML> <html> <head> </head> <body bgcolor=silver> <canvas id="myCanvas" width="600" height="400"></canvas> <img id="myHexagon" src="Images/Hexagon.png" style="display: none;"> <script> window.requestAnimFrame = (function(callback) { return window.requestAnimationFrame || window.webkitRequestAnimationFrame || window.mozRequestAnimationFrame || window.oRequestAnimationFrame || window.msRequestAnimationFrame || function(callback) { window.setTimeout(callback, 1000 / 60); }; })(); var animX = 0; var frameCounter = 0; var fps = 0; var time = new Date(); function animate() { var canvas = document.getElementById("myCanvas"); var context = canvas.getContext("2d"); context.clearRect(0, 0, canvas.width, canvas.height); animX += 1; if (animX == canvas.width) { animX = 0; } var image = document.getElementById("myHexagon"); context.drawImage(image, animX, 128); context.lineWidth=1; context.fillStyle="#000000"; context.lineStyle="#ffffff"; context.font="18px sans-serif"; context.fillText("fps: " + fps, 20, 20); ++frameCounter; var currentTime = new Date(); var elapsedTimeMS = currentTime - time; if (elapsedTimeMS >= 1000) { fps = frameCounter; frameCounter = 0; time = currentTime; } // request new frame requestAnimFrame(function() { animate(); }); } window.onload = function() { animate(); }; </script> </body> </html>

    Read the article

  • MDM Poised for Growth

    - by david.butler(at)oracle.com
    David Nixon, an Oracle colleague of mine, was doing some research on MDM the other day. He came up with some well founded insights that I thought I’d share with you. Gartner recently published a note asking “Should Organizations Using ERP 'Do' Master Data Management?”  It may seem a bit strange but that’s a question Gartner has been asked by a number of companies as organizations are beginning to understand the importance of data governance and data stewardship.  That’s because ERP Suites typically “focus on integrating their own applications within suites, but have little interest in making their suites interoperate with the applications or suites of other vendors.”  Therefore, Gartner is advising customers that “have deployed or plan to support multiple packaged application suites (even from the same vendor) that have different semantic data and/or process models” to add an MDM solution. And it appears that customers are taking note.  In a more recent note entitled “Search Analytics Trends: Master Data Management”, Gartner noted that MDM searches on gartner.com in November 2010 “were 300% higher than [in] May 2009, indicating the increased interest an importance that businesses are placing on MDM.”  Why the increased interest?  Moving towards a single version of the truth is a familiar theme, but customers are talking more about the underlying business value that this enables.  For example, businesses are talking about the need to fix master data before they can successfully move forward on SOA initiatives.  And the growing demands for compliance continue to be a major driver.  In short, companies are talking more about specific and tangible business value, and they are looking for help creating business cases for an MDM initiative. Why This Matters Gartner’s notes make three things clear.  First, MDM is poised for growth as organizations gain a greater understanding for it and the need they have.  Many are still sorting it out, but the demand is growing and is sure to rise.  Second, any organization with a heterogeneous computing environment should invest in MDM.  Even solutions from the same vendor may have different data models and could benefit from MDM.  But the key to growth, or which vendors will benefit the most from it, is the third and perhaps most critical point: companies need help with the business case for MDM. Oracle can help your organization build a compelling business case for MDM. We have seen our 1100+ MDM customers gain competitive advantages in a wide variety of implementations. Give us a ring.

    Read the article

  • Be Careful When Referencing SPList.Items

    - by Brian Jackett
    Be very careful how you reference your SPListItem objects through the SharePoint API.  I’ll say it again.  Be very careful how you reference your SPListItem objects through the SharePoint API.  Ok, now that you get the point that this will be a “learn from my mistakes and don’t do unsmart things like I did” post, let’s dig into what it was that I did poorly. Scenario     For the past year I’ve been building custom .Net applications that are hosted through SharePoint.  These application involve a number of SharePoint lists, external databases, custom web parts, and other SharePoint elements to provide functionality.  About two weeks ago I received a message from one of our end users that a custom application was performing slowly.  Specifically performance was slow when users were performing actions that interacted with the primary SharePoint list storing data for that app. The Problem     I took a copy of the production site into a dev environment to investigate the code that was executing.  After attaching the debugger and running through the code I quickly found pieces of code referencing SPListItem objects (like below) that were performing very poorly: SPListItem myItem = SPContext.Current.Web.Lists["List Name"].Items.GetItemById(value); // do updates on SPListItem retrieved     As it turns out the SPList I was referencing was fairly large at ~1000 items and weighing in over 150 MB.  You see the problem with my above code is that I retrieved the SPListItem by first (unnecessarily) going through the Items member of the list.  As I understand it, when doing so the executing code will attempt to resolve that entity and pull it from the database and into RAM (all 150 MB.)  This causes the equivalent of a 50 car pile up in terms of performance with a single update taking more than 15 seconds. The Solution     The solution is actually quite simple and I wish I had realized this during development.  Instead of going through the Items member it is possible to call GetItemById(…) directly on the SPList as in the example below: SPListItem myItem = SPContext.Current.Web.Lists["List Name"].GetItemById(value); // do updates on SPListItem retrieved     After making this simple change performance skyrocketed and updates were back to less than a second.   Conclusion     When given the option between two solutions, usually the simplest is the best solution.  In my scenario I was adding extra complexity going through the API the long way around to get to the objects I needed and it ended up hurting performance greatly.  Luckily we were able to find and resolve the performance issue in a relatively short amount of time.  Like I said at the beginning of the post, learn from my mistakes and hope it helps you.         -Frog Out   Image linked from http://www.freespirit.com/files/IMAGE/COVER/LARGE/BeCarefulSafe.jpg

    Read the article

  • Mathemagics - 3 consecutive number

    - by PointsToShare
    © 2011 By: Dov Trietsch. All rights reserved Three Consecutive numbers When I was young and handsome (OK, OK, just young), my father used to challenge us with riddles and tricks involving Logic, Math and general knowledge. Most of the time, at least after reaching the ripe age of 10, I would see thru his tricks in no time. This one is a bit more subtle. I had to think about it for close to an hour and then when I had the ‘AHA!’ effect, I could not understand why it had taken me so long. So here it is. You select a volunteer from the audience (or a shill, but that would be cheating!) and ask him to select three consecutive numbers, all of them 1 or 2 digits. So {1, 2, 3} would be good, albeit trivial set, as would {8, 9, 10} or {97, 98, 99} but not {99, 99, 100} (why?!). Now, using a calculator – and these days almost every phone has a built in calculator – he is to perform these steps: 1.      Select a single digit 2.      Multiply it by 3 and write it down 3.      Add the 3 consecutive numbers 4.      Add the number from step 2 5.      Multiply the sum by 67 6.      Now tell me the last 2 digits of the result and also the number you wrote down in step 2 I will tell you which numbers you selected. How do I do this? I’ll give you the mechanical answer, but because I like you to have the pleasure of an ‘AHA!’ effect, I will not really explain the ‘why’. So let’s you selected 30, 31, and 32 and also that your 3 multiple was 24, so here is what you get 30 + 31 + 32 = 93 93 + 24 = 117 117 x 67 = 7839, last 2 digits are 39, so you say “the last 2 digits are 39, and the other number is 24.” Now, I divide 24 by 3 getting 8. I subtract 8 from 39 and get 31. I then subtract 1 from this getting 30, and say: “You selected 30, 31, and 32.” This is the ‘how’. I leave the ‘why’ to you! That’s all folks! PS do you really want to know why? Post a feedback below. When 11 people or more will have asked for it, I’ll add a link to the full explanation.

    Read the article

  • DELL Inspiron 9400 SD Card Reader not working on 10.11 ubuntu

    - by Mario Martz
    Im new in linux, just installed Ubuntu 11.10 in a Dell Inspiron 9400, everything works fine with the exception of the SD card reader, everytime I insert a card the computer doesnt do anything, its like the SD card reader is not there. I did a $lspci and it shows the next drivers 03:01.0 FireWire (IEEE 1394): Ricoh Co Ltd R5C832 IEEE 1394 Controller 03:01.1 SD Host controller: Ricoh Co Ltd R5C822 SD/SDIO/MMC/MS/MSPro Host Adapter (rev 19) 03:01.2 System peripheral: Ricoh Co Ltd R5C592 Memory Stick Bus Host Adapter (rev 0a) 03:01.3 System peripheral: Ricoh Co Ltd xD-Picture Card Controller (rev 05) Everytime I insert a memory card, dmesg shows the next d status 0x600b00 [ 2687.227351] end_request: I/O error, dev mmcblk0, sector 64 [ 2687.229436] mmcblk0: error -110 sending read/write command, response 0x0, card status 0x600b00 [ 2687.229440] end_request: I/O error, dev mmcblk0, sector 65 [ 2687.230512] mmcblk0: error -110 sending read/write command, response 0x0, card status 0x600b00 [ 2687.230515] end_request: I/O error, dev mmcblk0, sector 66 [ 2687.231588] mmcblk0: error -110 sending read/write command, response 0x0, card status 0x600b00 [ 2687.231592] end_request: I/O error, dev mmcblk0, sector 67 [ 2687.232674] mmcblk0: error -110 sending read/write command, response 0x0, card status 0x600b00 [ 2687.232678] end_request: I/O error, dev mmcblk0, sector 68 [ 2687.234763] mmcblk0: error -110 sending read/write command, response 0x0, card status 0x600b00 [ 2687.234766] end_request: I/O error, dev mmcblk0, sector 69 [ 2687.236864] mmcblk0: error -110 sending read/write command, response 0x0, card status 0x600b00 [ 2687.236868] end_request: I/O error, dev mmcblk0, sector 70 [ 2687.238942] mmcblk0: error -110 sending read/write command, response 0x0, card status 0x600b00 [ 2687.238946] end_request: I/O error, dev mmcblk0, sector 71 [ 2687.238949] Buffer I/O error on device mmcblk0, logical block 8 [ 2687.241028] mmcblk0: retrying using single block read [ 2687.243104] mmcblk0: error -110 sending read/write command, response 0x0, card status 0x600b00 [ 2687.243108] end_request: I/O error, dev mmcblk0, sector 64 [ 2687.245212] mmcblk0: error -110 sending read/write command, response 0x0, card status 0x600b00 [ 2687.245215] end_request: I/O error, dev mmcblk0, sector 65 [ 2687.247298] mmcblk0: error -110 sending read/write command, response 0x0, card status 0x600b00 [ 2687.247302] end_request: I/O error, dev mmcblk0, sector 66 [ 2687.248389] mmcblk0: error -110 sending read/write command, response 0x0, card status 0x600b00 [ 2687.248393] end_request: I/O error, dev mmcblk0, sector 67 [ 2687.250476] mmcblk0: error -110 sending read/write command, response 0x0, card status 0x600b00 [ 2687.250480] end_request: I/O error, dev mmcblk0, sector 68 [ 2687.252617] mmcblk0: error -110 sending read/write command, response 0x0, card status 0x600b00 [ 2687.252621] end_request: I/O error, dev mmcblk0, sector 69 [ 2687.254737] mmcblk0: error -110 sending read/write command, response 0x0, card status 0x600b00 and more of the same but with different sector number Im using Kernel 3.0.0-12-generic By the way, when I was installing it and ubuntu asks about installation (If I want to install it along with windows or delete something or change the partitions of the HDD) if I go to the window to change the partition of the disc, linux detects the SD Card (if there's one inserted course). Any help with this it would be appreciated -sorry for my english Thank you

    Read the article

  • Partner Webcast – More out of ODA with DB Options - 19 July 2012

    - by Thanos
    The Simple, Reliable, Affordable Path to High-Availability Databases Critical business data needs to be available 24/7 for users and customers, but it can be a struggle to find the time and resources to build a highly available database system that’s reliable and affordable. That’s why Oracle created the new Oracle Database Appliance—a complete package of software, server, storage, and networking. The Oracle Database Appliance integrates the world’s most popular database - Oracle Database 11g  - with system software, servers, storage and networking in a single box. Business gets the benefit of a reliable, secure and highly available database to support applications and maintain continuity – as well as groundbreaking ease of use. But that is not all, with the support for all Oracle Database Options, Oracle Database Appliance can be the ideal solution for many use cases. The benefits?   Unmatched performance, reliability & security for your data that’s there when you need it – which is all the time. Fast installation, simple deployment, easy management. Out of the box. Significant cost savings & reduced risk and complexity compared to integrating all the elements yourself. Ongoing lower total cost of ownership with multiple automated support, detection & correction functions that also save you time.   Discover the Oracle Database Appliance Value Proposition and learn how to position and combine it with database options to capture new business and easily roll out solutions safely and with maximum cost efficiency. Agenda: Oracle Database& Engineered Systems Innovation. What’s the Oracle Database Appliance ? Oracle Database Appliance Value Proposition. Oracle Database Appliance with Database Options Oracle Database Appliance Partners Business Delivery Format This FREE online LIVE eSeminar will be delivered over the Web. Registrations received less than 24hours prior to start time may not receive confirmation to attend. Duration: 1 hour Register Now! For any questions please contact us at partner.imc-AT-beehiveonline.oracle-DOT-com Visit regularly our ISV Migration Center blog Or Follow us @oracleimc to learn more on Oracle Technologies as well as upcoming partner webcasts and events.

    Read the article

  • Good approach for hundreds of comsumers and big files

    - by ????? ???????
    I have several files (nearly 1GB each) with data. Data is a string line. I need to process each of these files with several hundreds of consumers. Each of these consumers does some processing that differs from others. Consumers do not write anywhere concurrently. They only need input string. After processing they update their local buffers. Consumers can easily be executed in parallel. Important: With one specific file each consumer has to process all lines (without skipping) in correct order (as they appear in file). The order of processing different files doesn't matter. Processing of a single line by one consumer is comparably fast. I expect less than 50 microseconds on Corei5. So now I'm looking for the good approach to this problem. This is going to be be a part of a .NET project, so please let's stick with .NET only (C# is preferable). I know about TPL and DataFlow. I guess that the most relevant would be BroadcastBlock. But i think that the problem here is that with each line I'll have to wait for all consumers to finish in order to post the new one. I guess that it would be not very efficient. I think that ideally situation would be something like this: One thread reads from file and writes to the buffer. Each consumer, when it is ready, reads the line from the buffer concurrently and processes it. The entry from the buffer shouldn't be deleted as one consumer reads it. It can be deleted only when all consumers have processed it. TPL schedules consumer threads itself. If one consumer outperforms the others, it shouldn't wait and can read more recent entries from the buffer. Am i right with this kind of approach? Whether yes or not, how can i implement the good solution? A bit was already discussed on StackOverflow: link

    Read the article

  • Access Control Service: Protocol and Token Transition

    - by Your DisplayName here!
    ACS v2 supports a number of protocols (WS-Federation, WS-Trust, OpenId, OAuth 2 / WRAP) and a number of token types (SWT, SAML 1.1/2.0) – see Vittorio’s Infographic here. Some protocols are designed for active client (WS-Trust, OAuth / WRAP) and some are designed for passive clients (WS-Federation, OpenID). One of the most obvious advantages of ACS is that it allows to transition between various protocols and token types. Once example would be using WS-Federation/SAML between your application and ACS to sign in with a Google account. Google is using OpenId and non-SAML tokens, but ACS transitions into WS-Federation and sends back a SAML token. This way you application only needs to understand a single protocol whereas ACS acts as a protocol bridge (see my ACS2 sample here). Another example would be transformation of a SAML token to a SWT. This is achieved by using the WRAP endpoint – you send a SAML token (from a registered identity provider) to ACS, and ACS turns it into a SWT token for the requested relying party, e.g. (using the WrapClient from Thinktecture.IdentityModel): [TestMethod] public void GetClaimsSamlToSwt() {     // get saml token from idp     var samlToken = Helper.GetSamlIdentityTokenForAcs();     // send to ACS for SWT converion     var swtToken = Helper.GetSimpleWebToken(samlToken);     var client = new HttpClient(Constants.BaseUri);     client.SetAccessToken(swtToken, WebClientTokenSchemes.OAuth);     // call REST service with SWT     var response = client.Get("wcf/client");     Assert.AreEqual<HttpStatusCode>(HttpStatusCode.OK, response.StatusCode); } There are more protocol transitions possible – but they are not so obvious. A popular example would be how to call a REST/SOAP service using e.g. a LiveId login. In the next post I will show you how to approach that scenario.

    Read the article

  • Should I choose Doctrine 2 or Propel 1.5/1.6, and why?

    - by Billy ONeal
    I'd like to hear from those who have used Doctrine 2 (or later) and Propel 1.5 (or later). Most comparisons between these two object relational mappers are based on old versions -- Doctrine 1 versus Propel 1.3/1.4, and both ORMs went through significant redesigns in their recent revisions. For example, most of the criticism of Propel seems to center around the "ModelName Peer" classes, which are deprecated in 1.5 in any case. Here's what I've accumulated so far (And I've tried to make this list as balanced as possible...): Propel Pros Extremely IDE friendly, because actual code is generated, instead of relying on PHP magic methods. This means IDE features like code completion are actually helpful. Fast (In terms of database usage -- no runtime introspection is done on the database) Clean migration between schema versions (at least in the 1.6 beta) Can generate PHP 5.3 models (i.e. namespaces) Easy to chain a lot of things into a single database query with things like useXxx methods. (See the "code completion" video above) Cons Requires an extra build step, namely building the model classes. Generated code needs rebuilt whenever Propel version is changed, a setting is changed, or the schema changes. This might be unintuitive to some and custom methods applied to the model are lost. (I think?) Some useful features (i.e. version behavior, schema migrations) are in beta status. Doctrine Pros More popular Doctrine Query Language can express potentially more complicated relationships between data than easily possible with Propel's ActiveRecord strategy. Easier to add reusable behaviors when compared with Propel. DocBlock based commenting for building the schema is embedded in the actual PHP instead of a separate XML file. Uses PHP 5.3 Namespaces everywhere Cons Requires learning an entirely new programming language (Doctrine Query Language) Implemented in terms of "magic methods" in several places, making IDE autocomplete worthless. Requires database introspection and thus is slightly slower than Propel by default; caching can remove this but the caching adds considerable complexity. Fewer behaviors are included in the core codebase. Several features Propel provides out of the box (such as Nested Set) are available only through extensions. Freakin' HUGE :) This I have gleaned though only through reading the documentation available for both tools -- I've not actually built anything yet. I'd like to hear from those who have used both tools though, to share their experience on pros/cons of each library, and what their recommendation is at this point :)

    Read the article

  • System Variables, Stored Procedures or Functions for Meta Data

    - by BuckWoody
    Whenever you want to know something about SQL Server’s configuration, whether that’s the Instance itself or a database, you have a few options. If you want to know “dynamic” data, such as how much memory or CPU is consumed or what a particular query is doing, you should be using the Dynamic Management Views (DMVs) that you can read about here: http://msdn.microsoft.com/en-us/library/ms188754.aspx  But if you’re looking for how much memory is installed on the server, the version of the Instance, the drive letters of the backups and so on, you have other choices. The first of these are system variables. You access these with a SELECT statement, and they are useful when you need a discrete value for use, say in another query or to put into a table. You can read more about those here: http://msdn.microsoft.com/en-us/library/ms173823.aspx You also have a few stored procedures you can use. These often bring back a lot more data, pre-formatted for the screen. You access these with the EXECUTE syntax. It is a bit more difficult to take the data they return and get a single value or place the results in another table, but it is possible. You can read more about those here: http://msdn.microsoft.com/en-us/library/ms187961.aspx Yet another option is to use a system function, which you access with a SELECT statement, which also brings back a discrete value that you can use in a test or to place in another table. You can read about those here: http://msdn.microsoft.com/en-us/library/ms187812.aspx  By the way, many of these constructs simply query from tables in the master or msdb databases for the Instance or the system tables in a user database. You can get much of the information there as well, and there are even system views in each database to show you the meta-data dealing with structure – more on that here: http://msdn.microsoft.com/en-us/library/ms186778.aspx  Some of these choices are the only way to get at a certain piece of data. But others overlap – you can use one or the other, they both come back with the same data. So, like many Microsoft products, you have multiple ways to do the same thing. And that’s OK – just research what each is used for and how it’s intended to be used, and you’ll be able to select (pun intended) the right choice. Share this post: email it! | bookmark it! | digg it! | reddit! | kick it! | live it!

    Read the article

< Previous Page | 468 469 470 471 472 473 474 475 476 477 478 479  | Next Page >