Re: Predefined nfx considered harmful David A. Wheeler 04 Sep 2012 20:12 UTC

Shiro Kawai:
> Thanks for the revision.   I'm going to implement it in Gauche
> and try it out.

Thanks so much!  Again, I believe experimentation is the only true test.  And in any case, there's no point in writing this stuff down unless it is *implemented*.

> On 'nfx': I think your concern about users relying on
> implementation-dependent extention inadvertently is reasonable.
> My $.02 is that you just say "if an implementations claims it's
> srfi-105, leave nfx unbound; if an implementation wants to put some
> default meaning on nfx, call it something other than srfi-105."


If I don't hear soon from someone, I plan to put in text like that.

> Of course a srfi-105 implementation may supply a library that binds
> nfx and let users load the library explicitly if they desire.

Absolutely!  But when a library is loaded, it's be obvious in the source code that there's a library dependency.  What worries me is silent dependencies.  I want to avoid the case where a {...}, used in a certain way, could create a quiet implementation dependency.  That's best avoided.

--- David A. Wheeler