Will,
> Actually, no it wouldn't. You still get the overhead of the error, plus > > any custom error handlers will be triggered regardless of the > > error_reporting setting which depending on the implementation of the > > error handler can be quite costly performance-wise. > > > > So what solution is there to this? Should it not be deprecated? This same > issue (is still happening) with register_globals for us, but it's > acceptable as it has been phased out. Should there be no deprecation > warning now and wait for a future release? > That's my suggestion. Officially deprecate it, but don't add E_DEPRECATED to it in 5.5. Update the documentation, and start a PR campaign to get off it. Then in NEXT add E_DEPRECATED and in NEXT+1 remove it. But I've said it a number of times, and it appears that I'm the only one who thinks that it's a good idea to take it slow and give users time to move off it... Anthony
