david reid wrote:
This discussion has just gone round and round in circles. How about I simply do what I started out doing - add an ssl socket implentation to apr-util. Once it's in, tested and working it should be easier to see if/when/how we then integrate the code into better/simpler usage models.
Exactly. Then if someone's itch is a 'unified' I/O approach, nobody would object either to an apr_io_read()/apr_io_write() that can handle all of the apr and apr-util objects with read/write methods (and poll and close etc etc).
