Testing Framework

Well, probably the easiest way to test a reliable UDP implementation is to modify it so that you can have the sender arbitrary drop outgoing packets and/or reorder them at certain rates.

If you don't want to muck up your code with those sorts of "features," you could write a small tool that could act as a proxy between two of your implementations and do the same thing.

Then, to test, you'd write two small programs that sent and received a large chunk of data between themselves and made sure it stayed correct and in order.

I think the proxy solution will cover testing the general working conditions of the implementation(basically, does it send/recieve reliably.. most of the time). The problems arise when I try to test the individual component parts of the system using assertions to ensure the 'most of the time' becomes an 'all of the time'. I can't expect to really know the exact state at which the system should be in if two components are operating independent of each other, and the proxy is adding a layer of uncertainty, so it removes any hope of installing package-wide testing. Is a testing framework on such a system too complex to consider such a precise level of validation?

Unit testing: Using knowledge about the implementation of a single module, you test the contracts of that module in the absence of any other module (except for necessary mock objects).

Functional testing: You make rules about the system behavior, and then you assert that behavior.

For example, if there is a rule that "eventually all realiable messages will make it through, or if not possible within 1 minute, report connection closed", then you can set up a scenario with 80% packet loss, and send a known sequence of 8 reliable messages. With a packet send rate of 10 per second, there is basically no chance that the messages won't make it through in a minute, so you can assert they make it through. Then set the packet loss to 100%, re-run the test, and assert that the system reports connection loss.