mark.umcp-cs@UDel-Relay@sri-unix (09/05/82)
From: Mark Weiser <mark.umcp-cs@UDel-Relay> Date: 1 Sep 82 22:38:12-EDT (Wed) TIOCIOANS and TIOCSIGNAL are for multiplexed files, a wonderful feature of Unix which allows a user program to pretend to be the tty driver. Berkeley says it will go away in 4.2, and I hope they provide something with the almost total transparency these things have.
pavel (09/10/82)
I really wish that the people who are bitching and moaning and cautioning Berkeley with regard to what 4.2bsd 'had better provide' would please read the 4.2bsd System Manual, freely available from Berkeley. The step from 4.1 to 4.2 is of a magnitude and \quality/ very different from anything else that Berkeley has yet released. The ARPA team has produced a very clean, very spartan kernel implementation, not just a feature-fest, as could be said of 4.1, for example. The new organisation is extremely flexible, extremely efficient, and very, very promising. I would feel comfortable staying with this kernel/system organisation for at least the forseeable future. Rather than try to convince you without giving you the background, let me just ask you to get and read that manual before you moan. 4.2bsd puts to rest, for me anyway, any questions about whose UNIX is the one to buy and/or support. The Unix Support Group at BTL may as well just start learning about 4.2; System III has been obsoleted. Pavel Curtis Cornell University Pavel.Cornell@Udel-Relay decvax!cornell!pavel
trt (09/12/82)
I would like discussion of 4.2bsd in unix-wizards, if not for Berkeley, then for other UNIX implementors. I have read a preliminary report on 4.2bsd. It just desribed what is in the new kernel without attempting to explain what the features were good for. Here are my own impressions of 4.2bsd: 1) It is ENORMOUS. Yes, it no longer has the creat(II) system call. But instead it has a sophistiated open(II). And it has sophisticated network support. And lots more, like real-time features and IPC features. And the race to make them *sufficiently* sophisticated has just started. My guess is the 4.2bsd kernel will have twice the code of 4.1bsd. 2) It has things that are neither necessary nor sufficient. a) kernel-level file locking, a subject which has generated many a flame. The locking is quite rudimentary--why bother? Was it required by DARPA? b) New "readv" (and "writev") system call, which replaces a sequence of read(II)s with a single "readv". Why? Efficiency? Indivisibilty? Couldn't there just be a "sysv(syscall, args, syscall, ..., 0)"? 3) It has things which are unproven, at least in the UNIX domain. a) The "dwrap" system call, which was vaguely described and which provides access to the networking code. b) Symbolic links. c) File names up to 255 characters long (and new directory file format). d) Many others (sorry, I can not remember them offhand). In summary, 4.2bsd appears to be an aggressive, innovative undertaking which provides a far bigger "feature-fest" than did 4.1bsd. Some new features sound wonderful, but only time will tell which are buggy, which are robust, and which are used. The system interface is so different from UNIX V7 that many would argue it is not a UNIX at all. Tom Truscott (duke!trt)