+972 52-548-6969
Building Frameworks: It is one day or three months
Quite often, people remark to me that I set unreasonable time estimates for work because I am somehow much better than anyone else. I strongly disagree with this statement. Both the qualitative assessment of my skills, and that this is an unreasonable timeframe.
There are several things that I am doing that I think that people are ignoring. One of them is not starting from zero, but starting already at high speed. But that isn't the major thing here. The major thing is that people fail to notice what I am giving those estimates very well defined things. You want to build an IoC container? It takes less than ten minutes to build a useful container. And I built an application on top of a container that wasn't much more complex than that.
That isn't a good IoC tool by any means, but it took very little time to build, and it works. If you click the link to see the container code, the container whose code you are reading is going to be involved in serving the code. It works and it fits to purpose.
That is a major thing to remember. A lot of the really useful concepts are trivially simple to build if what you need is a very simple. You can do that in a day.
And I include in this concepts such as OR/M, Aspect Orientation, IoC, Logging, Hot Code Swapping, Message Buses, Search Engines, Event Brokers and many other basic infrastructures.
The catch is that in a day, you get something that works, for the scenario that you need it for. If you want to take it to the next level, however, this requires a much higher degree of investment.
Coming back to the title of this post, you either do it in a day, or you dedicate three months for it.
Comments
A little bit OT. Some time ago in one of the ASP.NET MVC posts you've mentioned that you don't understand why there are so many sealed or non virtual classes in ASP.NET MVC. On last PDC Brad Abrams said that they've done that intentionally using the rule which can be expressed more or less this way: "open only those extension points which you have reasonable scenario to extend". (you can see a full PDC session at http://channel9.msdn.com/pdc2008/PC58/ ). Bad rule or bad use of rule ? Just curious your opinion...
Specific experience is also an important topic - I've built OR/Ms twice. The first one took me about a week to get fully functional, and the second one - half a day; and they had the exact same features. That's because the second time I know exactly where the pitfalls would be, even though it was a year later. I didn't stop to think for a sec, I just coded on auto-pilot. which is probably what you do when you build an IoC container, or something like that.
Comment preview