I've taken a first pass at editing the final section of the ReplayGain specification: Player requirements (http://wiki.hydrogenaudio.org/index.php?title=ReplayGain_specification#Player_requirements).
I have flagged two recommendations there that I am not sure match current RG practice:
1/ Recommendation to apply average RG value to a track without metadata. With the target level adjusted upwards to -14 dB compared to -20 dB in the original proposal, playing at unity gain in this case appears to be a tenable solution. It is certainly a simple solution. I think this should be the recommendation. The suggestion in the original proposal to use an average RG value does not appear to be fully baked.
2/ Recommendation for a hard limiter. Is there a player that takes this approach. Doing a good-sounding limiter is tricky and there's certainly not enough information given for a developer to get it right. I would propose removing this recommendation from the specification and adding a note indicating that it is an alternative clipping prevention approach.
I ended up simply softening the language on these two items.
Three fall-back strategies are discussed for dealing with missing gain metadata - using album when track is missing, using an average from other tracks and using unity gain. The averaging strategy is discussed in less detail than what appeared in the original proposal.
The hard limiter was left in the specification as a recommendation without significant details about how to implement it. The gain reduction clipping prevention strategy is discussed in detail including a new formula that calculates gain, pre-amp and clipping prevention together as one.
I also updated the example dialog to have an up-to-date graphical appearance and I tried to edit and clarify some of its contents.