An interesting post at audiofora (=r3mix.net) :
Why does LAME let you mess up in the first place?

Last time this discussion came up, the impression I got was that the LAME developers who read this forum are in favor of making one of the --alt-presets (probably either standard or fast standard) the default, but some internal LAME politics are preventing that from happening. But I agree, it definitely should happen at some point; the default LAME behavior is vastly inferior.

If the actual LAME developers aren't going to do it, we can at least work around the problem by having the various front-ends default to --alt-preset encoding. I believe EAC already does this.

Yep. I've thought of this before too. I wonder what possible 'internal politics' reason would stop this logical, and helpful, step from happening?

I suggest you guys all go post on the lame-dev list

In any way, there is a first step that should be done:

The ability to control the internals of alt-preset by switches (for advanced testers/ tuners/ developpers). This means that first some options should be implemented:

--psymodel gpsycho/nspsytune
--quantization 0...6
--quantization-short 0...6
--quantization-alt 0...6

--quantization-lower x
--ath-shape-modif x


It is not politic. The main point was that we can not remove already offered options (at least in 3.x).

But I do not see any real point preventing from defaulting to presets.
I do not think that you should flood lame-dev yet. First we should think carefully about pros and cons, and establish a list of needed steps to reach this goal.