TESTING

We can perform testing in many ways.

We would like to have testing which:
 - is highly automatic in performing
 - can test wide spectrum of modules and functins
 - should test behaviour with respect to original implementation
 - should test many possible code paths and conditions

This leads use to the case that we should have tests which are
automatically generated, and they if possible should be randomized.

They should be automatically run and compared with reference implementation.

For full automaticity we can run something on the server,
like Rhino or SpiderMonkey. We then can run it from script
with specific tests, then run also them in Erlang, and compare
using Erlang or some script.

We can eventually do this by spawning JS engine as Erlang port,
and send there commands.

We can also have webbrowser which will run them and upload
results using POST.

Other possibility is to imlement Erlang distributed protocol
or some kind of emulation so it will be transparent
to communicate beetwen the two nodes with normal Erlang
message passing.

Then it can be easly tested on both nodes. Randomized tests
can easly be runned on the server or client side,
and similary comparission can be performed both on server and client
(we can have bugs in comparission code/implementation,
or term decoding, or by incident, and the fact that we have
same data in memory, something could go wrong).

Most simple way will be to generate some tests on client (it will not
stress server), or on server (to have really good randomnes for sure :D),
then calculate both functions on both nodes (sending second one the needed
arguments and other parameters), then compare them on client
and show results. Client will also then report status of tests to the
server (with also failed tests cases for logging purpose).
If needed we can send all tests and their status and their results,
and server will optionally verify it again if it is ok.