Joseph Kowalski wrote:
> Joseph Kowalski wrote:
>> I seem to recall squashing a request to do that many years ago
>> (with great relish). Maybe twice.
> Speaking of "squashing", there have been multiple attempts to
> publicly expose these symbols. The rationale is that third party
> libraries could use the same trick - by binding to _read() rather
> than read(). That "trick" has merit, but I don't think it ever 
> happened.
> 
> (Bart, I think you were a proponent of this. Comments?)
> 
> - jek3
> 

I was and am a proponent of fixing the "surprise" factor that
occurs when application/library function (and data, shudder)
names intersect unexpectedly.  Originally, we did not have the
linker technology to do this w/o synonyms; the enhancements
made to the linker(s) in the intervening years (and the massive
cleanups of ON by Roger, Rod, and others) now allows us to use
direct binding to address this problem.  This is critical for long-term
stability of dynamic library interfaces; as we move to a more asynchronous
software delivery model & mechanism, removing the chance of
accidental interposition after the fact is even more critical.

I'm wondering if  we should be changing some default linking behaviors
when building shared libraries...

- Bart

-- 
Bart Smaalders                  Solaris Kernel Performance
barts at cyber.eng.sun.com              http://blogs.sun.com/barts
"You will contribute more with mercurial than with thunderbird."

Reply via email to