Architectural Improvement Request Formdoc: Fill & Download for Free

Super short versionNetKernel is a piece of software that will change the way you think about software.Online: http://netkernel.orgExecutive SummaryNetKernel is a dual-licensed advanced software environment from 1060 Research Ltd. in the UK. It grew out of some initial research at HP Labs in Cambridge with the goal of creating a software ecosystem that benefitted more directly from the multiprocessor, multicore CPUs that were beginning to emerge.NetKernel is a productive and scalable dual-licensed software environment that reflects characteristics that we have traditionally liked from Unix pipes and filters, REST and Service Oriented Architectures. Once developers get over the learning curve, it becomes straightforward to write efficient, scalable code that is easy to maintain and resilient in the face of changing business needs.Built on top of the Java VM, NetKernel allows developers to use whatever language suits their needs while isolating code that depends upon it from caring about the language implementation choices. Software is layered via logical connections that simultaneously shorten development time, improve runtime performance, provide dynamic systems that can leverage cloud technologies transparently and reduce the cost of maintenance.Personal anecdote: My first exposure to NetKernel was in 2004. I was working on an XML-messaging system that was running into some serious performance issues in the commercial environment that was being used. After a week of poking around with NetKernel, I was able to port the pipeline (20-30 Java-based steps) to NetKernel without changing any code, just how the pipeline was defined. Doing nothing else, we saw a 4x throughput increase on the same hardware.One step that we had never been able to accomplish with the expressive limitations of the other environment was to parallelize the extraction step. Once I had our pipeline expressed as a serially-connected series of steps, it took me ten minutes to parallelize the extraction step. As I learned more about the environment, if I was able to handle a throughput of X messages on a 1 CPU box, I'd get 2X on a 2 CPU box and 4x on a 4 CPU box.There is advanced systems research going on behind the scenes. It will reward the effort to learn with improvements to productivity, better use of capital expenditure, cheaper maintenance cost and exceptionally dynamic systems.More Detailed DescriptionIn an attempt to solve the mismatch between software that was being written at the time and the new hardware architectures, 1060 began from basic principles and created an environment that reflected the economic principles of the Web.The "kernel" in the name refers to the microkernel implemented on top of the Java Virtual Machine. It is a lightweight environment in which all activity is dispatched to the kernel asynchronously. This fundamental choice allows just about anything that runs on top of NetKernel to take advantage of additional CPUs and cores even if they were not written to do so in many case.Endpoints within NetKernel are deployed in modules and given Uniform Resource Identifiers (URIs), logical names that get resolved to the endpoint when a user requests it. These names define an addressing scheme to both data and the code that consumes and processes it.As an example, to request an XML file to be read in via a dynamic language, you might say something like:doc = context.source("res:/somedir/somefile.xml")To return the contents of the file as the result of your script being called, you would say something like:context.createResponseFrom(doc)If your module responded to the name "active:fetch-data", then a client elsewhere in the system could say:data = context.source("active:fetch-data")// ... do something with the dataThis "client" knows nothing about how the data is produced, it just knows that it can expect back some XML. If your module changes where this data comes from, the client will not be affected as long as you don't fundamentally break the expectations. This can be proximately accomplished by using interfaces in Java, but without some special bend-over-backwards shenanigans, you are going to be limited in how you can vary the form of the response.If you had an existing Java library that processed DOM Document class instances, you could use it and simply ask for the document as a DOM Document. NetKernel would take care of this process (known as 'transreption') for you. If you wanted it back as a String, a JDOM document, a byte array, a node set, etc., you can simply ask for it that way. This is a form of declarative content negotiation. As long as there is a converter that understands how to do the conversion, it can simply happen as needed. This makes it very easy to reuse existing libraries even if they don't have a common model and drastically reduces the amount of grunt code.This also makes it very easy to leverage languages like XSLT or XQuery that expect a nodeset view of the world even if the rest of your code uses Java objects. Something that might take hundreds of lines to accomplish w/ DOM Document object models could be accomplished in a few lines of XSLT or XQuery.All of this form conversion sounds like it would be expensive and unnecessary, but the good news is that, because everything in NetKernel talks through these explicit URI requests, we effectively get free architectural caching.Take for example the task of applying an XSLT transformation with a particular stylesheet. You would invoke the behavior by invoking the active:xslt engine with the parameters. If we wanted to apply the stylesheet to the results of the previous data request, we would issue a subrequest for:active:xslt+operator@res:/somedir/style.xsl+operand@active:fetch-dataAs expressed, you are expected to find that an arcane invocation, but I assure you, you quickly get used to it, but can also build up the requests with whatever higher level object abstractions you like. Behind the scenes, that it how it will be canonicalized.The benefits of the approach manifest themselves quickly:1) The active:fetch-data request can be cached if it is useful to do so. Usually NetKernel can figure this out automatically and keep track of what makes sense to. If you have special caching needs you have complete control of handling crazy custom policies or disabling it altogether. The point is, for free, if it doesn't need to do things, it doesn't.2) The reading in of the file in the satisfaction of the active:fetch-data request can also be cached. The style sheet request can as well. After this request is done, so can the application of the style sheet to the data returned from the active:fetch-data request. NetKernel is by no means *JUST* a caching engine, it just happens to be an exceptionally good one that optimizes based on actual runtime heuristics.3) The code that issues the above request will be largely shielded from changes of technology and implementation behind the scenes. The folks who built NetKernel recently noted that code that uses the XSLT engine (as above) will not have broken in the last ten years even though NetKernel itself has been rebuilt from the ground up twice in the interim. Code is stable. Layering is logical. Change is embraced. This is a beautiful orchestration/SOA/integration environment, but it is also an environment for building stable architectures that are expressed in terms of logical, business needs in ways that hide technological flux.4) The interpretation of the requests (such as active:fetch-data) are done within a context. That context can change at run time. In practice, that means that you can change where data and calculations are done dynamically even if the code has never been written to provide that kind of functionality. This effectively allows complete dynamic "gridification" of your entire infrastructure.5) The URI request mechanism is eminently interceptable which allows for aspect-oriented programming (AOP)-like concern modularization and weaving. Security, throttling, transactions, etc. are all isolated and changeable in phenomenally easy ways.Our tendency is to think in relational tropes... metaphors, similes, etc. We spatialize concepts based on how they relate to other things we grok. If we approach NetKernel with this mentality, we will attempt to characterize it as, perhaps, an n-tier "application server". It could just as easily be an orchestration environment, a SOA, a pipeline or some other thing. The reality is that it is all of these things. It is an environment for establishing architectures.This mini-essay is unlikely to have provided a satisfactory answer to the question but hopefully has intrigued you enough to pursue further investigation. The effort will be rewarded.

Industrial Plant and Robotics Automation Confederation - IPLRAC - A Good and Knowledgeable Place to Get Hands on Practical Training. IPLRAC Chennai deals with Distributed Middle wares in Detail.A distributed middle ware is a connectivity software designed for the interaction of software components in complex, distributed applications. During the 90’s, this technology has greatly evolved to provide interoperability and application portability in client/server architectures.A distributed middle ware, however, has emerged as an enabling technology also for distributed industrial robotic applications, requiring the ability to interconnect multiple and possible heterogeneous platforms while ensuring location transparency and extensibility. Among the most prominent actors in the distributed middle ware arena are Web Services and CORBA technologies.Service-oriented computing [Bichier et al., 2006] defines an architectural style whose goal is to achieve loose coupling among interacting software entities. In this context, Web Services [W3C, 2004][Motahari Nezad et al., 2006] provide a standard, simple and lightweight mechanisms for exchanging structured and typed information between servic es in a decentralized and distributed environment.The main goal of Web Services is to allow machine-to-machine interaction, whereas traditional Web applications are human-to human oriented. Relevant specifications and standards for Web Services include extensible Markup Language (XML), Simple Object Access Protocol (SOAP), the communication protocol,Web Service Description Language (WSDL), a machine-processable format to describe service interfaces, and Universal Description, Discovery and Integration (UDDI), centralized or distributed repositories listing service interfaces.We refer to the cited literature for the full development of these concepts. A Web Service contract describes provided functionalities in term of messages. By focusing solely on messages, the Web Service model is completely language, platform, and object modelagnostic.As long as the contract that explains service capabilities and message sequences and protocols it expects is honoured, the implementations of Web Services and service consumers can vary independently without affecting the application at the other end of the conversation.The use of Web Services is rapidly expanding driven by the need for application-to-application communication and interoperability. The question thus is whether Web Services are the right middleware for robotics as well. In our view, while the generality and reuse capabilities of Web Services are appealing for robotic applications as well, they fail to provide the performance guarantees for end-to-end system operation required in robotic applications.Several Distributed Object Computing (DOC) middleware standards are available, notably the Common Object Request Broker Architecture (CORBA) [OMG A, 2004] promoted by the Object Management Group (OMG), and Microsoft’s Distributed Component Object Model (DCOM). The distributed objects concept is an important paradigm, because it is relatively easy to hide distribution aspects behind an object’s interface. Furthermore, since an object can be virtually anything, it is also a powerful paradigm for building systems.DCOM’s major design goal appears to be providing improved functionality while staying compatible with previous versions incorporated in the early Windows systems. Unfortunately, DCOM has proven exceedingly complex, especially considering its proprietary nature. Today, DCOM has been largely superseded by its successor .NET, which however remains a proprietary solution mainly restricted to Microsoft operating systems.solution that provides similar mechanisms for transparently accessing remote distributed objects while overcoming the interoperability problems of .NET. CORBA is the result of an effort to provide a standard middleware platform to allow applications from many different software manufacturers to interoperate.CORBA allows smooth integration of heterogeneous software components. To ensure portability, reusability, and interoperability, CORBA architecture is based on the Object Request Broker (ORB), a fundamental component that behaves as a system bus, connecting objects operating in an arbitrary configuration

Below is my comparison between Atom & Sublime. Response format adapted from Sarath Nair's response to this question:Text Editor Age:Atom: ~ 2 years (see: Introducing Atom)Sublime Text: ~8 years (see: Sublime Blog | 1.0)Number of Users:Atom: ~ over 1 million (see: Atom Reaches One Million Active Users (03/28/2016)Sublime Text: ~10.86M control (see: Stats - Package Control)Packages/Plugins:Atom: 5164 as of 10/29/2016 (see: Atom packages page )Sublime Text: ~3883 as of 03/27/2016 (see: Stats - Package Control)Themes:Atom: 1,768 as of 11/23/2016 (see: Atom themes page)Sublime Text: ~ about 568 as of 10/27/2016 (see: Labels - Package Control). Data grabbed from the tags themes & syntaxKeybindings/Shortcuts:Atom Editor & Sublime Text share many of the same keyboard shortcuts.Development Cycle:Atom: core currently has 48 core contributors (see:Atom) Over the past few months Atom has been releasing updates on a consistent monthly basis for it's minor releases (Ex: 1.6 ->1.7 ...) . Atom also has a bleeding edge stable beta version for users, which has been ahead of the regular version by 1 minor release (see: Atom beta)Sublime Text: Number of known core contributors about < 3 (see: Sublime Text 3 Build 3103). Development cycles aren't as consistent as Atom. Sometimes release cycles for Sublime take several months (see: Sublime Blog)Extensibility/Hackability:Atom: plugins are created with HTML/CSS & Javascript (see: Tools of the Trade) In my opinion, this is one of the best features of Atom. Since just about everything is created with web technologies & developing for Atom is akin to creating mini web apps, you can literally almost create anything. Here are a few packages that just wouldn't be possible to implement in many other editors:hydrogen: Run code and get results inline using Jupyter kernels like IPython, IJulia, and iTorchgit-time-machine: Visually interact with git commit history for a fileplatformio-ide-terminal A terminal package for Atom, complete with themes and more. You can use bash/zsh/fish and other shells too.Companies are building their own IDE's & tools on top of AtomFacebook (nuclide), Jibo(jibo-sdk), platform io (PlatformIO), debuggers for multiple programming languages like javascript(node-debugger, atom-bugs), python(atom-python-debugger) & swift(swift-debugger), among many other languages.Sublime Text: plugins are created with Python (see: Sublime Text - Plugin Basics)linters, themes, code execution output is also available within Sublime. However, due to Sublime's design API & architecture, you are limited with creating packages that take advantage of the editors overall UI, package design & implementation.Documentation:Atom: open source, very well documented, contains visuals & consistently updated (see: Atom flight manual)Sublime Text: closed source & decent documentation (see:Sublime Text - Support), however the community has stepped up & created an alternative version which is comparable to Atom's docs (Sublime Text Unofficial Documentation)Application Speed:Atom & Sublime Benchmark (December 2016) - credits to XINHONG LIU (see: Sublime Text vs Visual Studio Code vs Atom Performance Test (Dec 2016))Atom: Noticeably slower than Sublime Text, but it’s improving steadily. Start up time takes a few seconds & may depend on the hardware of your machine, for small - medium size projects it feels quick, however for large projects or files, it tends to get sluggish, hang & often crash. *Update IMO: I’ve found Atom’s overal fuzzy search algorithm for file search & autocompleteto be faster & more reliable than Sublime. ❗Atom isn’t suited at the moment for opening very large files, in my opinion, (ex: log files > 5mb), but Sublime is good at this.Sublime Text: Sublime Text is faster than Atom in many areas, but not all. Everything from application startup time, opening large projects, global search within the editor [...] etc, is very quick.Future PredictionsAtom: is only going to get faster over time, given that speed is a priority for the Atom team. It's userbase is going to drastically increase as it's actively maintained & getting better with each release, because it's open source people are submitted bugs, forking, sending their pull requests to core & community packages. Given the aforementioned, the more people that continue to use Atom the better it will get.Sublime Text: the future is hazy, given that there isn't much transparency with the project given that's a commercial product. Release cycles for Sublime are infrequent & there simply is not as much contributors to the editor as Atom. However, I still have hope for Sublime.Pricing :Atom: Free & open sourceSublime Text: Commercial & never ending free trialPlatforms:Atom: Mac, Linux & WindowsSublime Text: Mac, Linux & WindowsAdditional Resources:Podcast: The Changelog #241: The Story of Atom with Nathan SoboAtom EditorWikipedia Page: Atom (text editor)Website: atom.ioBlog: Atom BlogPackages: Atom's packages websiteThemes: atomthemes.io or Atom's themes websiteTwitter Account: @AtomEditorGithub Account: github.com/atomForums: Atom DiscussionRecommended videos:Atom Editor - Overview and Top Packages RoundupAtom Editor Tutorial playlistSublime TextWikipedia Page: Sublime TextWebsite: sublimetext.comBlog: Sublime BlogTwitter: @sublimehqPackages & Themes: Package Control | The Sublime Text package managerForums: Sublime ForumRecommended videos:Sublime Text Tutorial playlist