Understanding Dead Reckoning - Hard DRexit versus Soft DRexit

Yes there was some other clause agreed that did that. However it doesn't matter if the whole thing has been sidelined

I know you are on a different mission but in general the Composition Codex must adapt to changes in the FIDE laws. It's our central sun and we will always need to clarify which light rays we accept and which we divert!

Our missions align on this point: I totally agree. However, the process has proved comically difficult and fraught

Cross posting is fun - hope there are no misunderstandings

OK I remember why I think that dead position rule needs be considered before we resort to the conventions. Here's an old problem to illustrate:

White to play and mate in 2. (A.Buchanan, RML mailing list 1/2001)

If we consider the castling convention then it will say that castling is ok anyway. But the whole point of the problem is that DP rule tells us that castling is definitively ok, without any need to resort to the tacky conventions.

There risks being a kind of "logical cook", where there are two ways to derive the state of the game. It seemed to me then (and still does) that the simplest, most accurate solution is to say that we only consider the conventions when there is unresolvable doubt about the position. That's what the castling & e.p. conventions were invented for: to resolve doubt. They are not retro conventions themselves, although they have been built upon for many retro compositions.

[first read next post] Yes, example. Great! Now make the same problem where the function of the castling move is replaced by an e.p. move - which of course is just as unprovable as the castling move except that we made a convention to downgrade it. Are we now suddenly sure that e.p. is legitimate because we must reject DP?

This issue goes to the core of my paradigm. (Active) FIDE laws only apply to proof games since these are individual games that could be played by 2 players according to these laws. DP is a FIDE law, not a convention and must therefore be applied to proof games.

A diagram represents the complete collection of proof games legitimately leading to that diagram. If you intend to make a move (any move) which is not valid in all proof games then you must apply the appropriate convention to get approval (which you will get for castling but not for e.p.). Next you drop all proof games which do not permit your intended move.

Finally, you apply DP evaluation to the proof games left in your collection, in casu only proof games with castling right (not permission) in the diagram and it gets you your #2.

So it is as you wrote: the castling & e.p. conventions were invented to resolve doubt - only not after you applied FIDE laws but before it. In the case of DP this is particularly obvious because DP looks at the future and you cannot confidently handle the future with unresolved doubts from the past - say psychologists

I am sure this rings a bell in your mind since I wrote this many times. It is hard though to tolerate concepts of a paradigm which is not your own. The brain tries to take care of its own consistency.

The usual addendum:

OK, I missed the top layer in your problem. It does not change my analysis of this type of postion but it does change the analysis of this particular position.

The castling right is indeed certain because it was already certain after execution of the last move - or the position would have been dead before it. Walking the same route as in the preceding post, the proof game analysis yields: there are only proof games for the diagram where white has castling right. Therefore no convention is triggered. That's all. The same would be true for a similar construction with e.p. if such a thing is possible.

This is a purely retro-analytical problem with the same outcome under any set of conventions! It is unusual in the sense that not many positions exist where castling right (right, not permission!) can be proved.

And it's a nice demonstration problem! Last move: ...KxRe3.

Addendum 2: 

Anselan: That's what the castling & e.p. conventions were invented for: to resolve doubt. They are not retro conventions themselves, although they have been built upon for many retro compositions.

Ari: Note that almost all miscellaneous conventions are about resolving proof game doubt. Doubt on who is on move, how many repetitions and 50M-moves have occurred, whether you may assume castling & e.p. right and what we should assume when the simpler assumptions collide (RS, PRA, AP). It's all proof game doubt addressed by retro-logical conventions.

The others address changes to the FIDE rules which we consider necessary to make compositions more functional such as yes/no on dead positions,yes/no on automatic draws and yes/no on having a 50M law. FIDE laws operate on known game state attributes and are non-retro beyond superficial inspection of the scoresheet but in compositions they often manifest in conjunction with retro-logical operations. You might say some FIDE laws are retro-sensitive.

And, finally, we now have the joke set of iffy conventions which depend on subjective human judgement and chess skills, and have no formalizable definitions.

@texaspete: 

It's hard to explain to a non-expert in this material. It's also hard to explain to an expert in this material  The solution is completely obvious (1. 0-0) but the issue is on how it should be justified. For instance, you know there is a convention that castling is permitted unless it can be disproved. Now assume that the convention was: castling is only permitted if provable (like the e.p. convention). Could white still checkmate in 2? The surprising answer is: YES! 

I will refrain from analyzing the remainder but you can imagine it contains a treacherous but completely lawful device!