This project is a dedicated event tracker solution for Kerkesix walking event. Hosted version is (or used to be) available at: http://satku.kerkesix.fi.

Project has been a technology testing platform for Tero Teelahti for over ten years. It has always had bleeding edge technology that has changed from year to year. The final version you see here is a simplified archive version that is easier to run locally on any platform, and does not need as many external dependencies. If this event would have continued, a rewrite to React + Redux would have been due (as it would fit perfectly on the event based approach at server side).

First version had a separate command line client that outputted bar code readings to a CSV file, which was then imported to the web site through UI. This of course did not handle any real time edge cases that well. Technology was ASP.NET WebForms and Canvas (which was very new at that time). Site was hosted at Kerkesix current web hosting, and database was MS SQL.

Admin console was super simple and used naive username/password login.

Analytics was done with Google Analytics.

Second version was a rewrite to ASP.NET MVC, which was not even released at that time. This version also changed database to Raven.NET, which was also brand new alpha-level document database at that time.

Third version was a rewrite to be hosted at Azure Web Sites, with Azure Table Storage as backend. At this time storage was used as a document storage. one row per one attendee. This was a must as Raven.NET was not easily hosted on Azure at that time.

This version lasted for a while, and was extended with registration, automatic email sending from the site etc.

Admin console used Google, Facebook and Microsoft account authentication.

This was the first version that had proper deployment pipeline with separate DEV, QA, and PROD staging areas. Deployments were automated during the years before the next rewrite.

Fourth version was the pinnacle of the whole site: it was a rewrite to Command Query Segregation Principle (CQRS) style application, using event sourcing as data storage style. This meant, that there was no state as is in the backend, just events that created the state when they were replayed on site start. Another very fancy change was the real time functionality: all users that had the site open shared the same state via WebSockets (or fallbacks via SignalR). This included client side persistent buffer queues for bad networks, building the state in two different ways in client and in server, and all other complexities.

This rewrite also introduced persistent pub/sub queues, that were used to gain transactionality behavior with non-transactional resources. Each incoming command would go through a complex set of cloud queue topics with retries, duplicate checking and the whole lot. At this time the site would have scaled to very high simultaneous users, which was naturally completely unnecessary as this site has max couple of hundred simultaneous users.

Finally, this version was a Knockout single page application, with all kinds of dynamic view loading capabilities. TypeScript was brought into play.

Azure's Application Insights was introduced as second analytics platform.

Self service registration was added.

During this versions lifetime MVC and Knockout versions were changed, and a separate command line tool was introduced as operating and debugging pure event data proved to be very hard. CLI tool also had features to upload old event's data from old format into the new.

The last version was a rewrite to .NET Core and ASP.NET Core to get proper cross platform support. As very many third party dependencies would not work Core, the site was dramatically simplified. Lots of stuff remained unchanged still, like the way Javascript is loaded, and the client side SPA framework. Next versions would have introduced WebPack loading and newer SPA framework.

In this release registration was also separated to be a separate site.

At the same time organizing the event had lost some of its glory and the future of the event is uncertain. Therefore the repository was cleaned up and archived to Github to public - unfortunately with all the history removed to keep the Git repo size in control and to protect the attendees as some versions had uploadable attendee data inside the repo.