What this contributes to my research into application sharing is the removal, or at least almost, of platform dependence of a software solution. I have my host machine a Ubuntu Linux installed and Windows XP in virtual box. Now if I use use a Linux application sharing solution, I can share the windows application seamlessly and makes things much easier in terms of user experience. I guess if you have your guest and host operating systems installed the other way around, you would share both windows and Linux, but your Linux app would run within a window. You would not have the same satisfaction.

Coming across real-time collaboration in Google docs and other web application, I have thought that the technology behind these is what I needed and you may find some more information in my earlier posts (Operational Transformation for Application Sharing’ Aug 2011). The technology used is Operational Transformation and has allot of nice synchronizing techniques.

I was thinking of making this technology keep two system synchronized on the window management system level or the operation system level so that it is generic and any application can be shared rather specific applications only. I have however later realized this might bare little fruit as the X Windows System is designed to do the sending of keyboard, mouse and display updates across a network.

Pondering on how I could integrate window systems level messages with Operational Transformation, I decided to go back to the basics and start the process from scratch. I was looking at having a software that would only intercept the keyboard and mouse events and send them across the network to another computer that would accept these and pass on to the operating system. I have found a solution doing exactly that which was actually mentioned by a colleague last year already. The solution is InputDirector which is a KVM (Keyboard, Video, Mouse switch) software. The software has an option of mirroring events across the master node and all slave nodes networked. There is also the software Synergy.

One of the things my supervisor also suggested for his idea was that the hardware and software installation of the two machines should be exactly the same. This could be resolved with use of virtual machines. This way the solution can run on any existing computer. The virtual machine image can be cloned with all the necessary software for the classroom already installed. Keeping the virtual machine copies synchronized one can look at virtual machine replica techniques.

As the underlying transport protocol used is TCP, the issue of whether all keyboard and mouse events will arrive and in correct order is no problem. After trying out this InputDirector having the mouse and keyboard events mirrored, I have however found that there are serious constancy issues. Mouse clicks may not have same affect due to perhaps window focus. In an examples on one machine a mouse click may initiate an event such a menu or program open while on another only the focus of the virtual machine window is obtained and so from this point any other event is pointless. The speed or response time of the application may also affect sync as the host operating system and hardware still affects performance of virtual machine. The sync may would also be broken if anybody on any of the shared machine use their mouse and cause lost of focus or closure of an open menu. These same sync problem would be experienced on operating systems and applications not running in a virtual machine.

One way of reducing or handling some of the above mentioned issues would be to lock events from any other machine which do not have control. In a classroom situation, students would need to participate by being able to get control also, so some floor control mechanism is needed as seen in conferencing systems. It would also be necessary for students to discuss and have control of their own machines independently of others and this would allow them make changes and all machines would not be in sync anymore. The use of snapshots can be looked to allow students to make snapshot before making changes and then returning to synchronized snapshot. Student would also need to share the work with the rest of a class to maybe share solutions to a given problem or other.

There are some things that need to be looked at in terms of more clarity on the above proposed solutions:

• Need to see synchronization of virtual machine images. Need to see if the changes such is reflected when resynchronizing virtual machines as application configuration; desktop changes; desktop icon changes; new files and folders created; start menu changes; etc
• Need to see if the virtual machine images can be updated while running. Also look at snapshot updating and transfer.

Other issues for investigating is the similarity of Operational Transformation documents and application sharing events or such.

Index Terms – Application Sharing, Desktop Sharing, Virtual Classroom, Collaboration

Paper in PDF format [A Classification of Application Sharing Tools for Use in a Virtual Classroom].

I have done some test shown in an earlier post (Testing Framework ’11). The handling of a full screen update is tested and also the handling of a video playback at 1 frame/sec and 5 frame/sec. Tools were selected for test for each using a different protocol and their settings kept at default. I had TightVNC, Windows Remote Desktop, TeamViewer and FreeNX. Only the NX solution could playback a video at 5 frames/second, the other tools could only show at 1 frame/sec. The NX solution can even playback a normal rate of 25 frames/sec smoothly.

There are attributes or settings that can impact performance of these solutions such as the screen resolution, compression algorithm used, color depth and display encoding. The interest of measuring is however not exact numbers, but in terms of user satisfaction of perceived performance. It should be known whether solution can work well over low-bandwidth or needs high bandwidth. Also to see performance of tools over high latency and unstable networks. Performance tests would therefore be done with solutions configured:

• for best response time and
• for best quality

So far as I understand is that it orignaly started with collaboration of text documents and the basic system would be based on insert and delete operation. It would also deal with consistency maintenance, locking, conflict resolution and other functions that arise in group editing. There are solutions for different types of application such CoWord, CoAutoCad and others that enable specific single user applications to be used in real-time collaboration.

These projects would build on and extends the Transparent Adaptation (TA) technology as noted from the CoAutoCad. The CoWord and CoPowerPoint projects may also be build on this. The TA architecture consists of three components.

• Single-user Application (SA) provides conventional single-user AutoCAD functionalities and interface features,  but has no knowledge about multi-user collaboration
• The Collaboration Adaptor (CA), which bridges the single-user application with the underlying GCE (see below). T CA component uses the SA API (Application Programming Interface) to augment the SA with collaboration capabilities without accessing the source code of the SA.
• Generic Collaboration Engine (GCE), which provides application-independent collaboration capabilities in consistency maintenance, concurrency control, group awareness, interaction control, etc. The GCE component can be shared by a wide range of  TA-based collaborative applications.

Here is a video on a presentation titled “issues and Experiences in Designing Real-time Collaborative Editing Systems” on YouTube.

I thought it would be a perfect solution to use in production a replicated architecture desktop sharing system for use in a virtual classroom with bandwidth constraints.

I have always thought that there is no way to know the inner design and workings of closed sourced software and even wrote a post (Benefits for Analyzing Closed Source Tools ’10) on this matter. I have therefore been leaning heavily towards free and open source solutions in my earlier posts on this journal.  In my first attempt of a writing a report on my findings I have noted this also (Application and Screen Sharing Solutions Investigation Phase 1 Summary Report ’10). Open source or free would still however be preferable, but not explicit in this stage of my understanding of my research.

I have however came to realize that the design and inner workings of these closed sourced solutions are available in the patents. These patents would then give a very clear and concise design and architecture of their solutions as per their functions. Many if not most times open source solutions do not keep proper design and architecture documents as detailed as in in patents. If you ask for a design documents from open source developers, they would refer you to the source code. Although the source code is a complete account to very fine detail of the software design, many people would be like me and not wish to spend many hours just to understand the architecture before deciding where to make changes, if there is a part of code to reuse and matters as such.

This concern of mine is on the background that I believe that all ideas arise from a network of other ideas. This might not be an issue as I get better understanding of incorporating patented ideas into research intend for open use. Adding to my concern mentioned in the beginning, I see some authors of papers that I have read being authors or designers of patent documents. It might lead to that the ideas used in open source systems would be as those of the patents.

There may be many projects to increase accessibility to the internet in Southern Africa, but the costs for these internet subscriptions for broadband connection are still high. Some ISPs have opted to charge per data and provide low bandwidth connection uncapped. With choice between uncapped low bandwidth and capped high bandwidth connection, the use of traditional application sharing systems remains a problem as the costs is high as these solutions do high volume of network data transfers. With these issues, the use of SaaS solution is an attractive option for introducing ICT teaching amongst the many advantages these systems provide.

The concept of Software as a Service (SaaS) has been growing and has been made easier by cloud computing and flexible hosting services. There are many advantages to it and companies and institutions are increasingly moving to adopt this technology around the world. The increasing connectivity of individuals to the Internet adds fuel to the growth of the adoption of cloud computing and Saas. Some of the popular of the many web-based applications are Google docs, Zoho, Etherpad and Office Live which provide similar functions of office suites such as Microsoft Office and many are even free. Although these applications do not have all the functionalities in traditional software such as a word processor, it is very ideal for introducing people in disadvantaged communities towards such applications and skills. In this paper we are more interested in those that have real-time collaboration feature as it needs to be used in a live virtual classroom, but the techniques suggested below can be applied to others.

The great thing about these applications is that they are run on the local machine as it is JavaScripts that are downloaded and the user enjoys same experience as with any other locally run application. These have the replicated design architecture of application sharing systems whereas the other popular used application sharing systems have the centralized design architecture. As such by their design the data being transmitted is minimal and makes it great for low bandwidth connections. Bandwidth is not an issue with these systems and can therefore support many simultaneous users with a cheap internet connection subscription. Voice is still needed, but there are solutions such as Skype for example that can work well over low bandwidths. The network latency, network stability would however is an issue for a good experience for all to see the changes to the applications in real-time. This paper does only address applications using AJAX and not those that work with binary objects such as Java Applets, ActiveX Objects and Flash Objects as used by some web based applications.

A major issue for use of these tools as is for teaching in live virtual classroom is that the toolbar would be problematic as the mouse click would not be visible to the viewers or the students in the live classroom. We suggest strategies to resolve this problem and the issues involved.  One of the solutions is to add mouse cursor tracking functionality to the web based application by adding local JavaScripts to be executed by the web browser. An example solution is to use an add-on to the Mozilla Firefox web browser called GreaseMonkey. The mouse clicks would then be broadcasted to the students or participants through another server to coordinate collaboration. There exists technologies for the administration and coordination of real-time collaboration between many users called Operational Transformation and one can download open source versions. Programming techniques may be the same for different web applications, but the scripts have to be adapted for each in terms of the toolbar positions and menu functions. This would be resolved by having the mouse pointer displayed when it is clicked and its position broadcasted to the other participants.

The uses of these systems do not need to be limited to teaching how to use a particular web based application. With the use of the imagination it can be used to teach any discipline such as even languages for example. The solution suggested only provides real-time teaching using application across network restriction and constrained in developing countries or other. It allows for the use of readily available resources with little prerequisite of a lot of technology competency for students.

One interesting thing that my supervisor mentioned is when he used the gotomeeting to do a meeting with people at the ITU him being here in SA, he could see the keyboard strokes in realtime. Why do I find this interesting? It made me thought more about the dominant factor for the performance of a tool  in terms of latency and this is discussed in another post.

The whole point of this post is the question of what is the benchmark for a good performance? It would otherwise be a matter of opinion or perception. Some paper by Adobe mentioned of an industry standard or 3 seconds delay in screen update, which is very slow. Is it possible or feasible to expect an application sharing system to have less than a second response time between when you press the keystroke and displaying it on the screen or mouse click and screen update as would be when using your PC?

There is also the issue of whose screen update do you measure? This might be of interest if one person is much closer to the server with lower latency than another.

I started this journal with a small introduction (Introduction, ’09), it may need to be expanded. The introduction page would also need to be updated at this point as I have a better understanding of the project than what I had in the beginning as expected.

In the start of my research I immediately started searching for application sharing tools and posted my experiences with these tools. I have started by posting about Skype and Chisimba Realtime Tools.  A follow-up post needs to be made as Skype has made changes and has integrated screen sharing into Skype whereas it was done through a plug-in called Unyte before (Skype Application Sharing Results, 24 Nov ’09; Skype Application Sharing Magically Works, 28 Nov ’09). I could not get the Chisimba realtime tools to work (Chisimba Framework Realtime Tools, 21 Nov ’09), but need to seriously take another look at it. It should allow the possibility of direct connection between Namibia and South Africa or any other SADC country. In a later posts, I have posted a list of popular screen sharing tools (List of Popular Screen Sharing Tools, 28 Nov ’09), but this list has grown and an update of this is to be posted later. A first of classification of these tools were also done (Application & Screen Sharing Tools Classification – Part 1, 5 Feb ’10) and an update of this classification in the later posts and a paper written for the 2011 SATNAC conference. Another attempt of the classification of this tools is also posted in my first draft of a report on application sharing tools (Application and Screen Sharing Solutions Investigation Phase 1 Summary Report, 3 April ’10).

I have discussed the replicated architecture in a few posts and some thoughts on use cases in the virtual classroom (Replicated Architecture Application Sharing Use Case – WireShark, 5 May ’10; Replicated Architecture Application Sharing Use Case – Audio or Video Playback, 10 May ’10; Replicated Architecture Application Sharing Use Case – VOIP SoftPhone, 10 May ’10). These have to be consolidated however with an update on understanding the project goal in a later post.  More also needs to be said on the centralized architecture with these use cases (???). An idea for converting a solution from one architecture to another was described (Centralized Architecture Application Sharing Use Case – WireShark, 31 May ’10), but this was in the line that I would do of developement in my project, which does not seem so. In the thought of development, the benefits of closed source tools was discussed (Benefits for Analyzing Closed Source Tools, 12 Marh ’10).

I have then decided to experience real life scenarios of online virtual classroom and have documented these experiences briefly in three different cases so far (Remote Desktop Experience with TeamViewer, 8 April ’11; NX Online Session with Skype – 18 April 2011, 18 April ’11; Free Webinar: ELEARNING CONTENT DEVELOPMENT and ROI FOR BUSINESS, 23 May ’11). More of these trials and experiences would be documented in later time.

There are two documents I have prepared, an Analysis Framework (Analysis Framework, 23 May ’11) and Testing Framework (Testing Framework, 23 May ’11). These documents would include many of the ideas and discussion of previous posts.

Notes would be added to post to update with new knowledge, make corrections or make it clearer. These would be noted will prefix “ADDITIONAL NOTE:”.

The results are are shown for:

• This is the amount of bytes send for the screen update
• The duration of updating
• Average bandwidth in mbps
• Average packets send per second
• Average packet size

The input/output graph for each the activity is also shown.

To download document, click here [Screen and Application Sharing Software Testing Framework]

Look at my comment on this post for my notes on this document. The document is far from perfect and is uploaded as is with the idea that it is part the learning process in this journey of my research.