This is the penultimate day of Tech Ed and I’m beginning to feel the burn. Long hours slaving over a hot notebook (not a metaphor) combined with early mornings have finally worn down the chiselled figure that is me.
Well all lies aside, this morning I found it difficult to identify exactly which seminars appealed to me. I finally settled on an initially intriguing session on tracking User Experience (UX to the cognoscenti). Why track User Experience, well assuming you care, the idea is to incrementally improve the usability of applications by monitoring unbiased user behaviour within a statistically significant number of user actions.
To do this, the developer needs to track user activity, preferably in a non-intrusive way. The real magic is in the analysis of the data. The main message was, “The reality of user activity will blow your mind”. The presenter, the self-styled, Supreme and Exalted Dictator for Life of Rolling Thunder Computing (He even has this on his business cards), David Platt, had developed his own framework to capture the necessary data. The Framework, admittedly, was developed specifically for local applications. But, to be fair, I found what I saw of the implementation to be both naive and far from state of the art.
The real innovation, was in the uses Platt made of the data. Novel approaches were used to infer user intentions, something that certainly can’t be tracked directly. As David pointed out, many attempts to include the user in the data capture process simply corrupt the results.
After my little venture into anonymized spyware, I ditched the general session as planned and secreted myself within the hands-on-lab until lunch. Where I practised, lambda expressions, Linq queries and Extension methods until lunch lured me away.
After lunch, I ended up having an interesting chat with some other delegates, until I was too late for Tess’s discussion on hardcore debugging with WinDbg and little more than an optimistic outlook. Actually, it was packed early and they were turning them away by the time I got there (I wonder is that because of my fabulous review yesterday, nah). Anyway, it’s repeated tomorrow and I’m going to be early, so help me.
Well, at a loose end I wandered into another seminar, this time about “Geneva” Microsoft’s new offering in the world of Federated Identity. Patently, something needs to be done with Federated Identity as every client wants it (“Single sign-on” for the technically impaired), but all the previous offerings have been pretty unpalatable and proprietary for the most part, to boot.
We all know that Microsoft love saving the day, and today was no exception. Mario Szpuszta outlined all the nifty accomplishments of project Geneva in a strong Italian accent. His accent may have been strong, but so were his vocabulary and his familiarity with the subject. Giving an entertaining session together with a great demo, I came away with a good understanding of what’s on offer.
Project “Geneva” really appears to comprise of a suite of different servers following Microsoft’s “claim based” architecture for security. The different, but complementary, servers seem to be to allow the framework to be employed in a vast array of differing environments, often with the absolute bare minimum of configuration. Of course, the cloud on the horizon was mentioned. And I think I can be forgiven for thinking that the “claim based” approach looks awfully like SAML assertions. Apparently, all of the necessary key/token swaps are via open source mechanisms to boot.
The next Interactive Session, “Team System Anti-Patterns” turned out to be a cracker. The first genuinely funny presenter of the event, Richard Hundhausen, packed them in. He rapidly ran through twenty or so Team System no-nos. More than one of which, I recognised as being a bad idea from personal experience. I had several homeresque moments during his delivery. Also mentioned, were the new TFS Power tools for 2008 just released which I will be obtaining, on the very instant!
Lastly the TFS dogfooding session left a bad taste in my mouth. Not because the session didn’t perform, it did. But what was delivered was my nightmare, a seminar for project managers. I had always actually wondered, how extremely large teams were manageable, when in my experience development teams become unwieldy around the 18~20 mark.
Boy, did I find out. Apparently, in extremely large teams, statistics become extremely important; the talented and very senior presenter Stephanie Saad, had numbers and charts for everything. It made my head spin. Rather like in quantum theory, extremely large numbers of developers make it necessary to use statistics to predict the behaviour of the system, with any degree of accuracy. I’m also sure that, just as with Heisenberg’s uncertainty principal, observing one single developer too closely will adversely affect the outcome of what’s being observed.