Sunday, July 26, 2009

Romialyo.NET SDK is public (with an ESB)

I've decided to publish the libraries I use for development in most of my projects. You can get it here, at github. It currently contains:
  • Several extensions to the .NET Framework I found myself re-coding many times.
  • An Enterprise Service Bus implementation, inspired in the good concepts from nServiceBus and RhinoESB.


Romialyo's ESB supports MSMQ and in-process message sending.

This ESB has become central to most of my projects so it was very important for me to have this code in a unique location and manage its versions properly.

Romialyo's ESB was started from the ground, trying to improve on nServiceBus code which I found hard to compile, extend and understand. The biggest stopper I found with nServiceBus was that it was too difficult for me to create an InProcess transport that will allow me to use the ESB as an Event Aggregator to communicate views in Smart Client applications.

In this blog post, Jeremy D. Miller exposes his conclusion about the "Event Aggregator = Service Bus" idea (braindump #12). I've been using a service bus with an events based in-process transport as the Event Aggregator in my applications for a while, and I must say I've found the idea very attractive.

Wednesday, July 15, 2009

TEAM's TimeTracker es realmente público ahora

Hoy he recibido una sorpresa: TEAM's TimeTracker pasa a formar parte de la base de software de Softpedia.

Siempre pensé que la aplicación era útil, sin duda lo ha sido para mi y para los Expectros que la utilizan a diario, pero tampoco hay dudas de que tiene demasiados errores.

Creo que esta noticia hará que le dedique un tiempo a pulirla un poco.

Thursday, July 9, 2009

Verificación automática de las reglas de negocio

Contexto:
- Empresa muy grande e informatizada.
- Gran volumen de datos.
- Lógica de negocio complejísima.
- La empresa depende en todos los aspectos de que los procesos informáticos funcionen correctamente.

Pregunta:
¿Es posible que todas las verificaciones de los procesos sean manuales, que haya muchísimas dependencias temporales (X debe ejecutarse después de Y) y funcionales entre todos los procesos, y que NO SE DERRUMBE TODO?

Al parecer sí. El coste es alto: muchísimos desarrolladores y muchísimo esfuerzo extra, pero funciona.