Threads

[tytso at mit.edu]

   Date: Wed, 01 Nov 2000 15:48:32 -0800
   From: Dan Kegel <dank at alumni.caltech.edu>

   That's a bit harsh.  Why should the LSB discourage apps from using pthreads?
   Is it because it's too hard to list the things that must be obeyed 
   to have a pthread app that is portable to both Linux and other OS's?  
   I must be missing something.

The argument was that if an LSB-complaint application ships with
dependencies on the current behaviour where it differs from Posix (For
example, the reason why getpid() returns different values for different
threads is because it's necessary to make the non-standard, non-Posix
signalling semantics work.), then if we later become more POSIX-like,
applications which depend on the old broken behaviour will lose.

   Is it just that I'm thinking at the source-code level, not at the
   installing-binary-packages level?

The problem is certainly worse at the ABI level, but even at the API
(source) level, there was a hesitancy to tell people to write
applications that depend on the current Linux behaviours.

						- Ted