> ...but if you want be sure things will work then you play by the specs 
> or take your chances. 

How "sure" do you want to be?  Just because it's a "standard" doesn't 
mean that all the bugs are out (yet).   That applies to bugs in the 
standard itself as well as bugs in implementations that claim to 
conform to the standard.

The risk of problems is probably somewhat proportional to the weight 
of the standard document (complexity) divided by the number of years 
it has been out (learning curve).