In part 1, I introduced a distributed programming environment called Global System Builder (GSB).  The IDE is an ASP.NET web application.  Another part of GSB is a Winforms application that runs on a remote computer that hosts an IronPython engine and communicates back to the web server using Windows Communication Foundation (WCF).  Hosted in the web browser is a source code editor with syntax highlighting and an interactive console – both of which communicate with the IronPython engine on the remote computer to execute Python code (and soon to be IronRuby using the DLR).

Also hosted in the web browser is a Virtual Earth control for locating remote computers, a Remote Desktop control to host a user session to the remote computer and a Smalltalk-like System or Class Browser for browsing classes (i.e. Types), member info, etc., that were loaded in the Winform app (i.e. GSB Service) on the remote computer.

One of the features I am implementing is the capability to extract code comments from any assembly that was loaded into the System Browser via the remote service (i.e. the Winforms application on the remote computer).  The idea is that as assemblies are loaded into the remote service, WCF communicates with the ASP.NET web server by passing up reflected assembly information (i.e. TypeInfo, MemberInfo, etc., using System.Reflection) that is then bound to a custom GridView control in the web browser, including code comments.

I implemented an ASP.NET AJAX hover menu that will display context sensitive menu items depending in which grid you are hovering over.  These menu items are helper functions (almost like macros) that will add the necessary IronPython (or IronRuby) code snippet to the console or code editor window, for example, to add an assembly references, stub out methods, insert DSL, etc.  For this post, I am discussing a way to view code comments that are part of an assembly, which when clicked in the hover menu, will display a panel with the code comment (plus parameter info, etc.) for the item hovered over, depending whether it is a type, method, prop, etc.  Think of it as a web-based version of Lutz Roeder’s cool Reflector, but in a Smalltalk like System Browser.

As an aside, I have one (major) design issue with the way assemblies are packaged out of Visual Studio as there is no way to include the code comments that you meticulously put in your source code files into the actual assembly itself.  It is packaged as a separate XML file.  In my mind, packaging the code comments in a separate file simply breaks the cardinal rule of encapsulation.  I wish Visual Studio at least gave you the option to embed your code comments directly into the assembly itself.

OK, so how to get at the code comments?  Searching through my vast libraries of reusable open source code on the internet, I came across this post and a nicely packaged class called XMLComments.  Only takes one line of code to implement and retrieve code comments. Perfect. Thanks Stephen Toub!

I used System.Windows.Forms.dll as the guinea pig as it is a whopping 5 megs in size and so is the companion XML code documentation file.  Everything seems to work ok, but a few observations.  It seems even in Microsoft’s own code documentation, only 50% of the +2200 Types in the assembly have a code comment.  Also, what about assemblies that don’t have the companion XML file and therefore no code comments at all?

There are a couple of choices, one is to extend the XMLComments class to derive some form of comments.  Ultimately, I would like to be able to build comments into the assembly (separate XML file or assembly) by using GSB’s source code editor and the necessary comments syntax to inject or attach comments to existing assemblies that don’t have code comments (loaded into the System Browser).

How important is this feature to the project?  Admittedly it is not a core feature, but at the same time, the point of the System Browser is to browse assemblies for reuse, or more importantly, for programming at a higher level of abstraction, so having code comments that came with the assembly or being able to attach your own code comments to the existing assembly, would be a good thing.  In my previous life in the electronics R&D industry, getting a component, like an IC, without the accompanying documentation with the package renders the component useless.  Looking at the Types list in the screenshot above, what do you think AmbientValueAttribute does?  Not sure what to make of our software development industry where source code documentation is either non-existent or in a form or place that is makes it virtually unusable.

Reusing assemblies to program at a higher level of abstraction is another topic and several posts in of itself.  There seems to be a lot of issues (angst?) in our industry about reusing other people’s source code or assemblies.  I do it all the time, both professionally (aka my day job) and in my little hobby here.  In fact the majority of GSB reuses several source code libraries and assemblies.  Without the many people that have taken the trouble to build the open source components that make up this application (aside from my own code), I would never be in the position I am today with a working application.

Anyway, I digress.  Displaying code comments is one feature of a growing list of features I collected after presenting GSB to a few potential users (i.e. dynamic programmers).  Based on the growing feature list, I still have a long way to go – not the 90% complete I alluded to in my previous post.  Also, I have decided to release GSB as open source when the DLR (with IronPython, IronRuby, Dynamic VB and Managed JScript) reaches RTM.  This is going to take a little bit of architectural refactoring (mostly the hosting namespace) to plug in the DLR engine and provide multi dynamic language support much like seen in this excellent video with John Lam and Jim Hugunin.  All good though as I have nothing but time, yes

January 30, 2008 Update - Global System Builder is available for download at: http://www.codeplex.com/gsb/Release/ProjectReleases.aspx?ReleaseId=10314

Official web site: http://globalsystembuilder.com

Oxymoron?  Could be, but all too real in the world of software start-ups in my opinion.  The following is a short story providing some lessons learned to software developers interested in starting up their own company.  Why?  I wish I had this advice when I started my own company in 2001, but it seemed difficult to find any material, specifically targeted for software developers that think they have a good product idea and want to bring it to market.

First let’s talk about product success.  My team and I created a product called Bridgewerx that used BizTalk Server as its integration engine.  BizTalk Server is a middleware product from Microsoft.  Whether you call it an Enterprise Service Bus or Internet Service Bus or Internet Data Router (which we called our product back in 2002 when no such term existed) or a SOA “type” product, or delivered as an SaaS or packaged as an Integration Appliance or… really does not matter cause these are mostly marketing words that basically mean the same thing:  loosely-coupled, message-based application integration server.  Purists will disagree, but I can make BizTalk meet any of these terms.

A saying is practice makes perfect.  When you perform similar tasks over and over again, generally speaking, you get good at it.  As a Systems Integrator, we designed and delivered over 30 application integration solutions to our customers over a 4 year time span.  We thought we were pretty smart about it, generalizing components and reusing them for the next job.  In fact we got so “practiced” that we were more or less assembling reusable components and the only customizations were customer specific adapters, business rules, schemas and maps. Regardless of the type of application integration work we did, these patterns occurred every time.  We became experts at developing BizTalk solutions using several design patterns and practices we had discovered and refined over this 4 year time period.

Then we thought, we could turn this into a product.  We could take our design patterns, best practices and embody them into a graphical design time tool that allows Business Analysts to specify application integrations.  The tool would capture all of the data (and metadata) necessary to fully describe an application integration scenario (even complicated ones).  The output of the designer tool would populate and configure a Visual Studio solution with several projects that contained source artifacts from our library of reusable components, (e.g. Orchestration patterns similar to C# generics, runtime message envelope, utilities, etc.), plus user defined schemas, maps, business rules and even custom assemblies.

Through scripting, we automated the build and packaging process so the final deliverable was a MSI that a customer could install on their target infrastructure, run the installer and voila, a completed application integration project, now ready for operation.

Anyone that has been involved in any size application integration project knows how difficult it is.  We called ourselves plumbers because it was a dirty and complicated job.  Application integration inherently deals mostly with distributed systems, not just a single application, therefore complexity goes up, along with the myriad of file formats per application integration.  If you have been there, you know what I am talking about.

So what?  Well, our product code generates “complete” application integration solutions from a design time tool output specification.  That’s the one sentence description.  This means the amount of highly skilled, intensive manual labor effort in designing and coding the solution is dramatically reduced.  At least by a factor of 10X, in some cases by 100X, and in the extreme cases where you are talking about very large scale application integration with +100 points of integration, it is not even feasible using traditional (i.e. manual labor) techniques.

We believe this to mean product success.  It successfully answers a difficult problem in our software development world in the domain of application integrations.  It provides at least a 10X improvement over current methods and tools. 

If you want to read more about the design of the product, you can click on the Bridgewerx category on the right side of my blog.  If you want to see a one hour movie demonstrating Bridgewerx, sponsored by Microsoft (thanks Jim!), you can see it here.  Finally, if you want to see a live demo, contact me at my email address at the bottom of the page.

Forrester research says that “Application Integration is the top priority for 2007.”  So we should be all set for business success, yes?  No.

Business failure.  As the President of the company for 4 years, a co-founder and co-creator of the product, I failed to realize the vision I had for my business.  Why?  It was not just one reason, but there were a few factors that occurred at key milestones in our company operations that really broke our company (figuratively and literally). 

One mistake I made was not getting the right people aboard our company to help us from a business perspective.  Involving business partners/advisors in a highly advanced technology business who have (very) limited understanding of software development and the target market is problematic.  Or even worse, business partners/advisors that were successful in the software business with one type of product and applying the same principles to a product and market that is completely different is disastrous, as was the case in my company.

My business partners had backgrounds in “traditional” software business/marketing, i.e. put the product in the sales channel and build up a base of customers.  On the surface this makes sense and were going with what they know.  However, our product, using terminology from Geoffery Moore’s, "Crossing the Chasm”, was an innovative and disruptive product that required very early adopters.  And from my perspective, once we had half dozen early adopters, the play was to sell our IP to a much larger player that could integrate our innovation into their product.  In our case, one play was to sell directly to Microsoft and have the innovation built into BizTalk Server.  Hard to believe no effort was to put into this when the company was operational.  Another play was to sell to a much larger System Integrator, like Avanade, who perform large scale application integrations, to give them a competitive advantage.  Yet another exit strategy was to take the innovation and apply it to another middleware or SOA product.  Based on my latest research, it appears we still have a market advantage of being the only vendor to code generate “complete” application integration solutions from a design time tool output specification.  Technically speaking, using Microsoft terminology, our design time tool is a graphical Domain Specific Language (DSL) and our code generation system is a Software Factory.  Innovative and disruptive?  Yes.  Easily understood?  No.

Lesson #1  Make sure your business partners and advisors fully understand your products marketplace and value proposition, especially if it is an innovative, disruptive product.

That was my vision and intent.  However, dear fellow software developers looking to bring your product to market, I made one fatal mistake.  I lost control of my company through share dilution.  In order to acquire the people that could realize the vision I had, required giving up shares as, of course, there was little money to pay large salaries.  Or at least so I thought.  Once you lose your controlling interest in the company, well, you know, other people take advantage of this for their own self interest and the companies vision and product changes direction, and not necessarily for the better.

Lesson #2  Keep a controlling interest in your own software company at all costs.

Another mistake I made was not keeping the professional services division of the company running to bring in cash flow.  It seems obvious enough that you cannot turn an SI into an ISV overnight.  But this was lost on me and my business partners/advisors, who were pushing just to raise capital.  We could have also acquired another professional services company that was similar to ours (Hi Mike and Jame!) to not only bring in additional pro services dollars, but also prime the pump for existing customers that may be interested in trying our innovation out. This would have reduced the dilution of our company before we even got started.  Unfortunately, our advisory board and other business executives did not understand this and was an opportunity missed and may have been the tipping point looking back in hindsight.

Lesson #3  Find alternate methods of funding, particularly if you are an SI that already has a revenue channel.

Speaking of funding, if you need to raise funds, ensure that you can raise enough.  Research shows that to fully capitalize a product as a SaaS requires $35 million to be raised over 5 to 7 years.  Here in the Great White North, we were able to raise $5 million tops.  If this were a car race, we did not even have a chance to qualify, let alone participate in the actual race.

Lesson #4  If you need to raise capital make sure you can raise enough.

Followed closely by:

Lesson #5  Ensure the people that are raising the money actually have a track record of raising that kind of money.

I know it sounds simple enough, but at the time…  At the time there was lots of bravado and speculation, which on the bottom line means absolutely nothing.  

The final lesson I have yet to figure out.  Based on my market research, our company has a product innovation that does something no other product does in our market space of application integration.  Market research shows that application integration is a top priority for 2007.  What up?  My guess is that we are yet another company that has fallen into the Chasm.

On a personal note, my hope is that in this short story provides a few tidbits of advice (and a little bit of hard earned wisdom) that other budding software development entrepreneurs will find useful, which is the reason I wrote this in the first place.  Secondly, am I bitter?  No, it was one of the most exciting and exhilarating business adventure that I have been on , Would I do it again, absolutely! Of course I would do it differently, no regrets.

For software developers wanting to unshackle themselves from the security blanket of the day job to develop their idea, vision, product.  If you can’t stop thinking about it… Carpe diem baby!

This is an open source software project that is highly experimental.  In fact, it might be just plain crazy.  Our industry is a buzz about Web 2.0, SOA, distributed computing, etc.  Yet, what tools do we have to support distributed programming?

What I mean by distributed programming is the ability to program multiple remote computers (i.e. servers) from within one IDE instance, or as I call it, a DPE – Distributed Programming Environment:

This ASP.NET AJAX DPE consists of the following parts labeled in red:

1)  An Object or Class Bowser, similar to Lutz Roeder's outstanding Reflector, except mine is laid out in the old school Smalltalk System Brower format.  Where are the assemblies IronMath and System.Windows.Forms coming from?

2)  This is a Remote Desktop Connection (RDC) that allows connections to multiple remote computers.  One tab per connection.

3)  This is a Windows Forms application (or service if you will) that runs on the remote computer(s) that allows you to add assemblies to be reflected.  It uses Windows Communication Foundation (WCF) to communicate with the web server hosting the DPE, which then displays the reflected assembly information in the browser window GridViews.  In addition, this Windows Forms application hosts IronPython (and soon the DLR) which also communicates with the web server via WCF so that IronPython code can be executed on the remote server(s).

4)  An IronPython source code editor which provides syntax highlighting, handling multiple source code files, etc.  Again, using WCF to communicate with the IronPython DLL that is running on the remote server(s).

5)   Is the result of running the IronPython code that is displayed in the editor window.

6)   To find remote computers, I use Virtual Earth’s map control to display remote computer locations.  Optionally, a MapPoint Web Service can be used to upload custom locations and attributes.

7)   Oh yeah, an interactive command line console is a must have.

The context of which remote computer I am programming is activated by which tab has been selected.  Meaning that when I click on any tab, any related context, like other dependant tabs and grids are switched too. I add distributed computers by adding tabs for each RDC session, a source code editor, and an interactive interpreter (i.e. console) per remote computer.

Think of the web server as a middle-tier to the WCF enabled remote services that host’s the code for reflecting assemblies and executing Python code.

As I mentioned earlier, I will be embedding the DLR which will add support for more dynamic languages.  Also note that the web-based console window is generic in the sense that it can be a console to a remote cmd window or PowerShell window or…  Same goes for the source code editor, its syntax highlighting also supports multiple languages.

Like I say, it will either be useful or just plain crazy.  That’s why I call it Global System Builder

January 30, 2008 Update - Global System Builder is available for download at: http://www.codeplex.com/gsb/Release/ProjectReleases.aspx?ReleaseId=10314

Official web site: http://globalsystembuilder.com

Previously, I discussed a web-based source code editor for IronPython, which is part of a larger application that I am working on.  Most of the editor is complete, just working out the details on handling multiple source code files.  Today we will drive IronPythons’ interactive interpreter, using a web-based console, which is also part of my application.

Using the console, we can whip up a simple Windows forms application in seconds:

And here is the output:

The console is based on two components, one is Sam Stephenson’s Prototype JavaScript framework and the console GUI is brought to you by Jeff Lindsay’s Joshua, which I modified to work with IronPython.

Running a web console over the internet presents a few challenges.  Everywhere you read, AJAX is the sh!t.  I use a ton of it in the app I am building.  A stands for Asynchronous and therefore, as some have stated, synchronous over HTTP is bad form mate.  Well, in the case of a console application, I think a synchronous call is one (only?) way to make it “really” work:

//synchronous call to the server passing the IP string command and returning the response  

function getFile(url, passData)

{

    if (window.XMLHttpRequest)

    {

        AJAX = new XMLHttpRequest();

    }   

    else

    {

        AJAX = new ActiveXObject("Microsoft.XMLHTTP");

    }

    if (AJAX)

    {

        AJAX.open("POST",url,false);

        AJAX.setRequestHeader("Content-type", "text/xml");

        AJAX.send(passData);

        return AJAX.responseText;

    }

    else

    {

        return false;

    }

}

One nit pick with the ASP.NET AJAX framework is that you can’t do synchronous.  I would hope that the framework would support this option as a) other frameworks provide it and b) I want it man!  The point of a framework is to provide options so that a variety of similar, but different business requirements can be met with “one” AJAX framework.  I am already using 3 AJAX frameworks as each one brings a unique piece of functionality that I require.  Who knows, maybe it will be more by the time I am finished.

If you look at the console closely, you will see one small GUI  issue.  When parsing interactive = true with Python, visually, the next line of Python code should reside beside the “…” and not on the next line beside the command prompt (i.e. >>>).  Also note that the 4 spaces indent after the def statement does not work at the moment.

Aside from a couple of GUI issues, and a missing blinking cursor, there is one last feature to be implemented and then the console will be complete.

The astute reader may notice that the Windows application is running in a Terminal Services session, which is being hosted in my web browser. Did I remote into the web server? No.  On the client computer? No.  Where is the Windows application running?

January 30, 2008 Update - Global System Builder is available for download at: http://www.codeplex.com/gsb/Release/ProjectReleases.aspx?ReleaseId=10314

Official web site: http://globalsystembuilder.com

Why web-based?  Well, the IronPython source code editor I am working on is one piece of a larger web-based, open source application that I am building and will become clear in future posts as to the reasons why it needs to be web based.  

In the meantime, I thought I would write about some of the available components that provide real-time syntax highlighting and other features one might want in a web-based source code editor.  After extensive research, there are 3 web-based source code editors that I would recommend.

The screen shot above comes from a portion of my web-application that I am building.  I am using Christophe Dolivet fantastic EditArea.  EditArea is written in JavaScript and really is quite outstanding given the number of features it has, particularly that it is free (released under LGPL license).

One of my main requirements is that it can work with ASP.NET AJAX (a free framework) and can be embedded in a TabControl as you see in the screen shot.  I will discuss the web-based IronPython interactive console in another post. 

I am really impressed with what Christophe has done, not only is EditArea fully featured, but the JavaScript source code is extremely well written using prototypes.  Further, the documentation is excellent – you would think this is commercial based source code editor.  EditArea supports multiple languages, both the software itself and the syntax highlighting it supports, for example, in my case English and Python.

Another real-time syntax highlighting source code editor I have come across is CodePress by Fernando M.A.d.S.  It is also written in JavaScript and covered by LGPL.  I really liked this editor as well.  While it is not as full featured as EditArea, it does a very good job.  It also supports many languages, but not Python at the moment.  I see it is on the todo list.  While I could get it to run in a bare bones ASP.NET AJAX application, I could not get it to run embedded in a tab control.  However, I think the limitation may be my JavaScript coding skills (I am a C# programmer hooked on Intellisense) and not necessarily CodePress itself.

Syntax Highlighter by Marijn Haverbeke is another web based, real time syntax highlighter.  It is released under a BSD like license.  It works quite well, but not as fully featured as the other two editors.  However, it is quite fast and seamless.  Dos not support Python syntax highlighting, but as in the case of CodePress, would be fairly easy to implement.  What I found really interesting was Marijn’s story on how it was designed.  Fascinating Captain!

Thanks to Christophe, Fernando and Marijn for creating and developing these real-time syntax highlighting source code editors.  It has saved me a lot of time not only from building these from scratch, but also proves a theory I have about software industrialization.  That is, it has been done before and I don’t need to reinvent the wheel.  Some people call this mashups, others reusable software or open source, or…  No matter what you call it, I am thankful that so many people are willing to share their hard work.

When will it be released?  Not sure as there still is lots of work to be done and can only work on it part time.  I wish it was my full-time job!

January 30, 2008 Update - Global System Builder is available for download at: http://www.codeplex.com/gsb/Release/ProjectReleases.aspx?ReleaseId=10314

Official web site: http://globalsystembuilder.com

Years ago I lived in a Smalltalk world and never knew how good I had it.  As Larry O’Brien says, “Smalltalk has enough proponents so you're probably at least aware that it's browser and persistent workspace are life-altering (if you aren't, check out James Robertson's series of screencasts).

Unfortunately, in the Great White North, not too many Smalltalk jobs were available and in fact it really boiled down to programming Windows or Unix.  And somehow I found myself in the Windows world using Visual Studio and Visual Basic.  In some respects, VB was almost Smalltalk like – I loved the interpreter of stepping through code, finding a mistake, fixing it on the spot, setting next statement to run a few lines back and stepping through and keep on going.  I found myself enjoying the rhythm of it and how productive I was.  That was the thrill for me.

Then came C# which is a statically typed language, but what really got me was the rote of write code, compile it, fix compile errors, then compile it again, fix runtime errors, compile it again, step through, note the error, stop debugging, edit code, compile it again, etc.  As you can imagine from the VB world, the most common error I got was, “are you missing a cast?”  Also, my productivity just was nowhere near as fast.

A year or so ago, I came across IronPython.  What I loved immediately was the interactive console that interpreted code on the fly.  I type in my line of code and execute it with immediate results.  For me, I work best like this.  It allows me to experiment with CLR types and once I have it figured, I can then cut n paste into a code file and build up my application.

What is really exciting to me is that the DLR is intended to support a variety of dynamic languages, such as IronRuby by John Lam and others… yes, even Smalltalk.  Way to go Peter Fisk!