But people do write GCC-dependent code. In fact, they will even write GCC version dependent code, to the tune of "this will not compile correctly with GCC 4.x.

But people do write GCC-dependent code. In fact, they will even write GCC version dependent code, to the tune of "this will not compile correctly with GCC 4.x.x, you will need to use 2.95.26.2.3.5.1" And given that the complaint was specifically levelled against Ruby, for the most part, it seems to be making one specific point: If your language is specified by the existence of a single reference implementation, and you keep changing that reference implementation, then that's equivalent to the language continually changing, which is a fair criticism, I think, particularly when it breaks backwards compatibility or continually deprecates features used in production code (requiring an endless cycle of rewrites, or maintenance of some older branch of the implementation). This sort of behaviour would certainly strike me as "annoying", but that's essentially what you get when you commit yourself to using a product which makes no attempts to standardise, and which lacks large commercial customers who would place pressure on the software vendor to keep backwards-compatibility as a priority.