[gnugo-devel] unused code

[Arend Bayer]

The amount of code in the engine that is not (yet) used (by default) is
growing. Currently, we have at least the following:

 - life.c
 - half of readconnect.c
 - --experimental-semeai
 - --experimental-owl
 - --standard-connections
 - pattern based tactical reading
 - tree-based replacement of DFA
 - surround.c
 - oracle.c
 - a couple of smaller things (4 komaster schemes, hashing with
   fullboard fall-back, an interface to automatic tuning of move
   ordering in reading.c, probably more...)

Some of them haven't had much attention for quite a while, others might
become used by default soon.

It's certainly one of the advantages of a free software project by
several contributors instead of a single developer that we can try
very divergent approaches and then select the ones that work well.

However, it also comes with the down-side of making maintenance more
difficult, and making it a little more difficult to read through the
code. (I wonder how many people start looking at life.c when they want
to start understanding GNU Go's life and death reading.)

To reduce this penalty somewhat, I'd suggest:
1. (obvious) keeping the default-off code as separate as possible
2. When it would be a burden to accustom a patch to some code
that hasn't been maintained for a while, the patch can simply break it.
This is probably better sorted out once someone picks up working on
that code.

And if we agree that some code is obsolete and unlikely to be re-used
again, I think we shouldn't hesitate actually removing it. We can keep
the patches on the web in case someone wants to restart using them.

Arend