The Freak Parade

The pace of change in the business world is accelerating at a rather alarming rate, as you may have noticed. The mighty Internet is compressing time, geographical distance and even our attention spans with ruthless disregard for our spiritual well being. As the pace of change increases the ability of any given organization to navigate skillfully in such frantic waters won’t remain a mere competitive advantage but will soon become a matter of survival, if it isn’t already.

Gone are the days of blissfully sailing across the open sea gleefully screaming "Fire!" one minute and hitting the deck the next as your competitor’s logo-imprinted cannon balls whiz overhead. We must continue the battle,  of course, but the open sea has been replaced by the thin edge of an epic vortex that keeps growing larger, spinning faster. When everyone is always talking about increasing "Business Agility," whether it be through implementing a well-governed SOA or by calling upon the power of Satan, what they are really talking about is finding some way to keep from sliding down the side of the whirlpool and into the gaping maw of a giant squid.

Melodrama and nautical fantasy aside, the operating environment of the modern business is changing. Businesses are becoming increasingly sensitive to their environments as "Business Intelligence" - the buzz word that refers to actually making some use of the torrents of data flowing through our IT systems - begins to permeate our organizations.

Originally Business Intelligence (transactional data processed and presented for analysis) was employed primarily at the executive level to allow decision makers to analyze the past in order to make predictions (and therefore decisions) about the future. The current trend, however, is to push this kind of processed, aggregated and integrated data deeper and deeper into the organization, sometimes right down the the line level worker. This is being called "Operational Business Intelligence." Operational Business Intelligence is not at all strategic - it is intended to provide fast, objective feedback to operational units in order to improve real time or near real time decision making. Well, this kind of business intelligence isn’t your granddaddy’s business intelligence - it’s value is derived directly from its immediacy, its freshness and its narrow relevance. It isn’t a map for the Generals to study while they sip whiskey and play with plastic figurines. Rather, it is the static filled screams of "OH MY GOD THEY’RE COMING THROUGH THE FUCKING WIRE - I NEED AIR COVER NOW!" bellowing from the field lieutenants radio. It does little good to see such data on a report a day or two (or even an hour or two in some cases) after it is generated. This kind of data shouldn’t even be called analytics; it is direct feedback from living business processes and the good old fashioned ETL based nightly batch process that does some pretty math and coarse grained aggregations and takes forever to process just doesn’t cut the mustard. That kind of business intelligence it is too general and too damned old to be useful for making in the moment operational decisions.

Business Activity Monitoring (BAM) to the rescue! The acronym and the term were invented by some highfalutin Gartner analysts awhile back ago and refer to business intelligence systems which aggregate, correlate, and report (usually via live dashboards and other more immediate alert mechanisms) real time or near real time business data as it is generated by the transactional systems that run the business. Additionally, BAM systems will often employ a rules engine to detect actionable scenarios in the incoming event feeds and either initiate an automated response to such scenarios or alert an appropriate human who can take some appropriate action.

As you can probably imagine, the architecture and design of a BAM system must be dramatically different from that of a batch based, ETL powered system. Furthermore, simply populating the data warehouse isn’t enough - BAM systems deal with real time business events and data feeds and therefore must be prepared to detect and act upon (even if simply via some kind of alert or dashboard dashboard alarm) various business conditions. This kind of task is something data warehouses are ill equipped to address.

Rather than processing data at some interval, BAM solutions monitor business events and aggregate, process and act on those events as they are occurring. This style of processing lends itself almost perfectly to an Event Driven Architecture, or EDA. In an EDA distributed systems communicate with each other indirectly by raising "events" whenever some interesting state change occurs in the domain. These events are usually just messages placed on a Message Bus, or ESB, to which any other system also connected to the message bus can subscribe. At any given time, there may be hundreds or thousands or millions of "business events" flowing through an organizations message bus, providing a perfect opportunity for a BAM system to tap in and do its groovy business.

An Event Driven Architecture or Message Bus are not strictly required for a BAM system to work. BizTalk, for example, provides a BAM API which it expects LOB systems (or BizTalk orchestrations) to call into at various stages of processing to directly inform the core BizTalk BAM database of any interesting business activity. Somewhat more robustly, some of the enterprise BAM solutions provide a wide array of adapters that can allow transactional systems to publish events directly to the BAM software. I don’t think it would be fair to call this kind of arrangement an Event Driven Architecture; even though events are being piped directly out of transactional systems and into the BAM system, the event notifications are essentially a private conversation between each transactional system and the BAM system.

Even so, just because you can have BAM without an EDA and an EDA without BAM, the two technologies (or architectures) are perfectly suited for one another. BAM provides a layer of intelligence and analysis over real time business events, and an EDA provides a conduit for transactional systems to publish their business events to the broader enterprise in blissful ignorance of who else may be watching, including the BAM system. An EDA also provides a completely natural way for the BAM system to distribute information or initiate action when it detects the need to do so - it can just raise its own derived events, to which a dashboard, human notification system, etc. can happily subscribe. Put the two together and BAM! You have a closed feedback loop that will enable your operational managers and line level decision makers to successfully keep your Ship of Commerce out of the tentacles of a giant squid, and maybe even sink a few enemy galleons along the way.

Well, all this is just fascinating stuff, but if you aren’t a big enterprisey company with deep pockets, you will be pretty disappointed when you get to the BAM store with your fist full of change and your head full of high hopes. Most of the existing BAM vendors have been swallowed up by the infrastructure giants like IBM, Progress Software (Sonic ESB) and the like. Probably top notch solutions, but I wouldn’t know because I work for an SMB. Their salespeople generally won’t even acknowledge that companies like ours exist. The one free standing BAM vendor I could find, Systar, also prefers to sell to the big fish, or, as their "business development" guy put it, "Tier 1 Companies who pay licensing fees that start at a quarter of a million dollars". I’m not 100% sure what a Tier 1 company is, but based on the fact that he tried to convince me we didn’t really need a BAM solution, I’m guessing we’re not one of them :)  So, as usual, that leaves us SMB’s alone at the bar sitting next to a Microsoft solution, in this case BizTalk, trying to decide if we should just head home alone and take care of business ourselves, or down a few more shots of Tequila, heave a deep, forlorn sigh of resignation, turn to BizTalk and say "How you doin’?"

I should have more to say about that soon as we’ll be implementing a BAM/EDA solution over the next few months. Will a bag-over-its-head be enough to stomach an ongoing relationship with the BizTalk BAM product, or will we be able to stitch something together from open source tools and good old fashioned grit, piss and vinegar? Only The Shadow knows…

So this crazy metaphysical debate about how or why you (or your team) should learn TDD rages on, reaching a fevered philosophical pitch. There is a heavy drumbeat coming out of the ALT.NET encampment that is becoming increasingly difficult to ignore, and we’re finding ourselves torn on how to react. Do we have a moral imperative to march along? Is it enough to tap our feet under our desks and keep doing business as usual? Or should we tear off our clothes, paint our faces with the ALT.NET war paint, and take our place by the campfire? Ian Cooper recently pointed out that what is at question here is not TDD but the imperative for us, as an industry, to accept change as a fundamental aspect of our practice.

Change is nothing new. In fact, change is all there is. We all attempt to resist it in one way or another, but it is a useless effort, like trying to stand still in a rushing river. There is no such thing as "resisting" change - the best you can do is to pretend it doesn’t exist, but you can’t stand still. Even though you may not alter your behavior, your environment will change around you and your relationship to your environment will change despite your stubbornness, it can’t be helped. Whether change is good or bad is a meaningless question, and the question of whether to accept it or ignore it isn’t really a question of "self improvement" but a question of how much you will enjoy your profession and your life.

So while the call to change coming from ALT.NET is extraordinarily valuable, the relentless emphasis on the tangible business results of some particular change, such as a guaranteed suite of regression rests resulting from a move to TDD, creates a limiting environment in which to accept it meaningfully and realize its full value. The only way to accept change is to experience it with an open mind and feel its value for yourself. An emphasis on quantitative results is absolutely necessary for arousing interest in some new technique or methodology, but it should not be the only emphasis. Once interest is aroused the emphasis should immediately shift from the expected benefits which, at this stage only reinforces preconceptions and inhibits the learning process, to remaining observant, alert and curious about the actual experience and actual results the technique or methodology provide.

TDD is a perfect example because everybody is always talking about its objective benefits, such as a safety net for major refactorings, enabling collective code ownership, cleaner API’s resulting from testable API’s, but the real value of TDD is a complex mix of all of these things and more. You can listen with your jaw agape to  someone extolling its benefits for days at a time but until you actually start to practice it you can’t really have any understanding of it at all. This is true with everything, but with TDD the effect is very dramatic, and therefore very instructive. People are always saying "it’s about design, not testing" but then they go on to describe benefits that are often mostly about testing. This is because the design aspect of TDD is something you will feel when you try it with an open mind, it is extremely difficult to articulate. The results are not so hard to articulate, such as decreased coupling or increased isolation, but an understanding of why these benefits result from TDD is more visceral than intellectual.

People talk about learning curves as if they were impediments to obtaining the bountiful rewards of some shiny new methodology, but the learning curve is exactly what produces the rewards. The learning curve IS the primary value. The moment the learning curve flattens out is the moment you will find a new methodology to replace the old one. Ian said in his post that change results from the pain of a process outweighing our natural resistance to change, but I completely disagree. I believe change is the fundamental and most basic aspect of our professional lives, the very reason we enjoy what we do as much as we do, and that methodologies evolve not at the last possible moment when we can’t stand the agony any longer but at the first possible moment, the very moment we suspect our current practices cannot evolve any further. A work environment where nothing changes, where technologies are frozen at some arbitrary point in time and previously established development practices have taken on the aura of biblical law are most developers worst nightmare.

Yet, as Udi Dahan points out, we often feel we are not at liberty to go off experimenting with change at the expense of pressing business realities. And this is true to a large extent, and I agree with his conclusions - TDD or any other new process should be introduced into an environment responsibly, ensuring an appropriate level of safety for the business as well as the team, or the results could well end in disaster. The danger, though, is to use that as an excuse to avoid accepting change altogether. While it is suicide to ram some new methodology into the gears of a running machine, a more sinister form of suicide is to neglect change entirely because no time ever feels like the right time. Consciously and constantly adapting to change should be built into the foundation of your business culture and your development culture. This is the responsibility of the management and line level developers alike. If your company is not interested in continuously evolving along with the software industry and therefore does not provide a safe (if controlled) environment for experimenting with and adopting new technologies, practices and methodologies then you should find another job. Conversely if a developer on your team is not interested in adapting to change then you should transition them off your team, either out of the company or into a support role.

Finally, all this highlights for me the need to emphasize that business results are not why we come to work every day. You may like to believe that you invest yourself as deeply as you do in your career because you want to contribute to the success of your business, or provide the best possible experience for your users, or streamline business processes, and these things mean something, but you know and I know that those reasons aren’t the real reason you show up at your keyboard. You love your job because it is mysterious and fascinating and full of creative possibilities and interesting logical puzzles which give you immense pleasure to solve and explore. You love to create something out of nothing. Fortunately, this kind of enjoyment usually aligns, or can be made to align, quite well with business objectives. But the point is motivation - yes developers want to be efficient, yes developers want to deliver maximum business value, but even more than that they want to maximize their creative possibilities and they want to enjoy the time they spend in front of their computers. So if you want to sell TDD, Agile, IoC, or whatever it is, don’t forget to sell that aspect of things as well as the objective, tangible efficiency or quality gains.

My team recently attended an Open Space conference, Open Agile 2008, which is a sort of new age, ad-hoc, self organizing meeting of passionate people. An Open Space conference has a number of principals, but the one that really caught my attention was the commandment to "Be Prepared To Be Surprised." What is meant is this: put aside your pre-conceptions and expectations and expect to learn something new at any given moment, and in unexpected ways. This applies equally well to a new technology or methodology such as TDD: "You will come away with full test suites that will dramatically improve the quality of your software, your designs will have lower coupling and a cleaner API and fewer defects. Oh, and be prepared to be surprised."

Recently I’ve been doing most of my web browsing using Chrome, because it is new and fast and I generally don’t use a lot of plug-ins. Currently I have about 20 tabs open in the browser because I’m doing research on a few topics at once. Probably like most people, when I find some content I know I’ll want to read, I continue my further explorations in a new tab expecting to go through all the tabs later to read what I want to read or to catalog my findings using LaterLoop or Delicious. This works really well, except that every once in awhile something runs afoul in a particular web page and poof goes the browser. In IE 7, well, sucks to be me when that happens. My tabs were almost never restored. In FireFox 3 an unexpected crash did restore tabs upon reload, but the browser crashed a *lot* and starting a new instance and waiting for the tabs to load all at once is slow and rather painful. Worse, the web page that caused the crash would  be usually be marked for restore, so very often after a crash choosing to restore the previous session would result in a new crash, in an endless loop until I gave up on restoring the session or managed to get to the offending tab with a mad dash of my mouse and close it before it loaded the poisoned part of the page.

Well, those days of agony and uncertainty are gone with the introduction of "process-per-tab" browsing found in Chrome. I believe the newest version if IE will also support this model. Anyway, my computer suddenly slowed to a crawl, and task manager showed Chrome as the culprit. But not Chrome in general, just a tab in chrome. Actually, not a tab in chrome either, a window inside a tab in chrome that was hosting a flash animation, which had run amuck. Chrome actually has its own task manger, so the solution to my problem? Right click on the Chrome tool bar, choose task manager, look for excess processor usage, kill said process - and all is well. All 20 tabs breathed a collective sigh of relief, and I un-wet my pants and kissed my monitor. The damage? A portion of one of the web pages I had open showed a broken plugin graphic and a nice little notice at the top of the page said something like "A flash plug-in has crashed." A single tear rolled down my cheek for the poor, lonely flash animation I was forced to kill, but many more lives were saved. Thus is the price of progress.

A while ago I did a survey of some the CMS solutions available for the .NET platform. While there are currently a very large number of both commercial and open source CMS solutions available, the one that really caught my attention wasn’t available at the time for public review. Now it is not only available for download in either source or binary format, but is backed by what at first blush appears to be pretty extensive documentation via a dedicated wiki.

I’ll be taking a closer look at this over the next few weeks, but my general feeling is that this product will add tremendous value to the community, despite the many existing CMS products already available, and I am very excited to kick the tires.

Project Links

* Main Product Page: http://www.sensenet.hu/engine.aspx

* CodePlex site: http://www.codeplex.com/sensenet

* Documentation Wiki: http://wiki.sensenet.hu/index.php?title=Main_Page

* Discussion Forum: http://forum.sensenet.hu/

Sample Sites

Additionally, they have two case studies already of full blown, production sites built on their new system. The case studies give a brief overview of the projects and link to the live sites. Both of the case study web sites are Hungarian, though, so the experience is a bit awkward (unless you can read Hungarian, of course), but still gives you a flavor for the CMS. [Update] - the Invitel site has an English version here: english.invitel.hu