.NET or WinAPI

The biggest requirement is that it should be as small as possible.

If this requirement is legal, it pretty much forces you to go to the win32 API level. But I really doubt this claim. These days, multiple MB downloads are not what they used to be, and there are ways to mitigate the problems if most of your user base is still in dial-up mode. Comparing this with the increased complexity (and therefore time) for development for Win32 and .Net can make a lot of sense.

What percentage of Windows might not have .NET 2.0 on their computers?

It is difficult to get real penetration numbers and I have seen stats above> 90% and only 20%. So it all comes down to what you want to believe. If this is really that big deal you can design for mono and use it for runtime static linking.

Also, I've heard that developing this small application in WinAPI will take more effort.

Okay, imagine you have to build a car.

• Go to the auto parts catalog, you order the correct number of valves, pistons, gears, fan, radiator, etc. for the engine. You order all the lubricants, teeth, bearings, nuts and bolts needed to assemble the gearbox. You order headlights, bulbs for lanterns, therefore for burst lamps, turn signals, reverse lights.

You order enough steel to weld the chassis, order some seats, some kind of fabric for their upholstery, gear, pedals, switches for lights, doors, glass for windows, and finally a rearview mirror.

Everything will come and you have to put it all together. The chassis is as strong as the strongest weld, and you have a gigantic challenge to put it all together and the room for error is huge. It’s only after you put it all together that you realize you didn’t order wheel hubs or tires.

It is Win32, windows api.

2. You enter the catalog of auto parts, you order a chassis with seats, lights, wheels, etc. You order the engine and gearbox. Three points come. You still have to tweak the motor and hook up the gear (not trivial tasks), but this is a much simpler proposal.

This is the .net framework.

If the project is bigger than a remote controlled car for a small child, grab a distribution and use .net

Definitely use "native code" if it should be kept to a minimum. However, it is much more complicated, especially if you are not from software.

Maybe it's better to outsource this work?

Writing your own (unmanaged) code that calls WinAPI will give you a smaller footprint with no .NET dependencies.

But if you want to reduce deployment of the .NET environment, have a look at the .NET client profile . All .net assemblies you normally use in a 28MB area. (With a 200kb boot block that only installs things that are not already on the client machine)

See also this blogpost .

Is "as little as possible". The most important requirement? By following this route, you will have fewer dependencies (and therefore fewer deployable components), but your development costs and time will increase.

I'm going to go to a limb and say that most Windows PCs already have the .NET platform installed (either through Windows Update or pre-installed). If you are developing against XP or Vista, this will certainly be the case.

But more importantly, is the size of the final packet as important as the time-to-value? It will be much easier for you to find developers who can quickly write good code in .NET. WinAPI development can be tedious, slow, and time-consuming to debug. This does not mean that WinAPI developers are not very good at what they do, but .NET is an advancement in software development that allows you to focus more on business issues and less on the technical area.

If you are the prime minister or project manager, you may need to do a cost analysis to determine what really matters in your decision.