greg woods said:
   [ On Tue, November 15, 1994 at 12:02:37 (EST), Gordon W. Ross wrote: ]
   > Subject: How to support clone devices (i.e. /dev/tcp)
   >
   > Ugh!  That's really unfortunate.  I guess they couldn't figure out:
   >
   >   cc -o the_objs.o -Bstatic -lX11 ... -Bdynamic -lsocket -lnsl
   
   You've got to be kidding.  Anyone who's even tried it before knows very
   well that you have to stay as far as possible from "-lsocket" and
   especially from "-lbsd" on almost all SysVr4 ports.  They are buggy to
   no end, and the socket layer on top of STREAMS is known to be rather
   inefficient in many ports.  It isn't really safe to do anything but
   stick to the native stuff.

in what way is the socket library broken ?  what else do you suggest
people do ?  write code for both sockets and tli ?  i certainly won't
be changing any of my networking code to use either.  or is there
something else i'm unaware of ?

yes, the ucb library is majorly broken in svr4 (all of them, perhaps
except dgux...which isn't much like `plain' svr4 anyway, and doesn't
have a /usr/ucblib) .. but it's `hidden' away in /usr/ucblib ..
   
   Not to mention that the STREAMS interface is far "nicer" (perhaps more
   elegant would be one way to describe it) than the socket interface, at
   least IMHO, and that of many others.

interface to what?  STREAMS and sockets are a different level.  and
i'd agree with the "elegant" comment, but not with the "nicer" one.

.. this has moved away from tech-kern's charter i believe ..

.mrg.