> Simply put, Unix programs rarly are written with threads. They should be.
An application program written with threads has to deal with most of the
issues that we have to suffer with kernel programming (synchronisation,
memory coherency, non deterministic behaviour, limited stack size, etc...)
Such programs are painfull to maintain, especially when you haven't
written it yourself. They are poorly portable and may encounter weird
performance problems depending on the plat-form (not only UNIX)
implementation of threads, if, by chance, you have been able to port it.
This let me think that writing an application with threads is the worst
approach you can have if you have any portability requirements or if you
will have to make it evolve and to maintain it in the long term. The
event-driven approach is a lot more simple, and has _less_ overhead since
you need far _less_ synchronisation in order to make things right. The
issue is that UNIX kernels and some other O/Ses donnot provide a system
interface smart enough to handle event (message, IO, etc ..) waiting and
sending in a coherent way, and to prevent deadlock/livelock situations.
UNIX (orignal one (ones??)) has been the poorest O/S that has been ever
designed and implemented, in my opinion. But it had the vertu to be
simple/simplistic. This simplicity allowed many people to understand it
and this made it popular. The evilness has been, that lots of incompetent
idiots felt they were great architect since they understood UNIX basics,
but they were actually not so. As result, numerous UNIX-called O/Ses are
born, going mostly in erratic directions. All this extra-burden, didn't
help to redesign properly the UNIX system interface so that it will
_really_ fit needs of what application programs need in _real_ life.
(I speak about pre-1991 UNIX systems obviously. ;-) )
Useless to flame, if you disagree, since I am not going to change my mind
on this topic.;)
Regards,
Gérard.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/