Robert Haas <robertmhaas@gmail.com> writes: > ... I guess I'd prefer #2 to #2.5, #2.5 to #3, and #3 to #1. > I really don't like #1 much - I think I'd almost rather do nothing.
FWIW, that's about my evaluation of the alternatives as well. I fear that #1 would get a lot of pushback. If we think that something like "LATERAL ROWS FROM STRICT" is worth having on its own merits, then doing #2.5 seems worthwhile to me, but otherwise I'm just as happy with #2. David J. seems to feel that throwing an error (as in #2.5) rather than silently behaving incompatibly (as in #2) is important, but I'm not convinced. In a green field I think we'd prefer #2 over #2.5, so I'd rather go that direction.
I suspect the decision to error or not is a one or two line change in whatever form the final patch takes. It seems like approach #2 is acceptable on a theoretical level which implies there is no desire to make the existing LCM behavior available post-patch.
Assuming it is simple then everyone will have a chance to make their opinion known on whether the 2.0 or 2.5 variation is preferable for the final commit. If a decision needs to be made sooner due to a design decision I'd hope the author of the patch would make that known so we can bring this to resolution at that point instead.