Good, if Van Jacobson and Dennis Ritchie say that STREAMS are wrong,
be so :-) I never said that it would be good to replace sockets
with STREAMS in any case. Though, it is apparently correct for character
devices (as I know SCO still uses clists for ttys and STREAMS only for network),
and it is arguable for network interfaces. Yes, SVR4 STREAMS are not effective,
but it does not mean that they are deemed to be slow.
Alan, who inhibited to use pbufs with STREAMS?
: However, most System V programs don't use Streams directly, but rather
: they use TLI. And it's not very hard to emulate TLI on top of BSD
: sockets interface.
I thought the same thing when I started to play with TIRPC.
I changed my opinion when was experienced enough.
Now I am about to start company to assure people to make TLI standard
interface to kernel and emulate BSD sockets on the top of TLI :-).
It is joke, but:
Do you know that:
-TLI can pass IP options (and other things that required by RFCs)
with every datagram.
-TLI local transports has intrinsic credentials passing capability.
Every programmer who tried make something useful with UNIX datagram
sockets will vote for TLI.
-a lot of more things (f.e. selective accept, data with SYN, expedited
data handling)
Just look at SVR4 docs, it is exciting reading.
BTW I never read POSIX socket api draft. Is it possible to get
it from somewhere?
Resume:
- TLI conforms to IPv4 RFC requirements, BSD sockets are not.
- BSD sockets seem not to be deemed to conform to IPv6 requirements.
Looking at the last draft of IPv6 socket interface I am crying...
apparently, its authors decided to preserve the worst features
of BSD sockets.
Alexey Kuznetsov.