[comp.sys.amiga] Porting DNET to SYS V & Amiga uucp/news

gmd@pulsar.UUCP (George MacDonald) (08/11/89)

 Re: DNET for SYS V

In article <907@corpane.UUCP> sparks@corpane.UUCP (John Sparks) writes:

   Many Amiga/Unix users will appreciate it if DNET on SYS V is possible.
   Does it/ will it exist?

DNET utilizes the socket functions first provided in the Berkeley 4.2 
distribution of unix. These C language functions have been ported by a 
number of vendors to their system V based platforms. This is usually done 
when a vendor is interested in providing networking capabilities(TCP/IP, 
NFS ...). The system V branch of unix lacked good support for networking
and remote file sharing until SV.3 when support was added for streams and 
RFS(remote file sharing). 
The programmers library providing for this is called TLI(Transport Layer 
Interface), and is very similar to the socket function calls. In fact it is 
almost identical, so a port to Sys5.3 or greater looks to be relatively 
straight forward. Of course Sys 5.4(Out to developers now) will provide 
both sets of functions so it then becomes a moot point.

Porting DNET to anything before 5.3 without the socket library will be
a chore. The major problem in emulating the socket functions will be 
providing multiplexed input on a single channel(serial line). This needs
to be done in the kernel to do it properly, hence a device driver is needed.
I am not familiar with DNET internals, and not sure which address domain it 
uses (UDP, INET ...) or which socket communication style(stream or datagram).
If DNET uses UDP and stream sockets then emulation using named pipes seems
at least possible. If DNET uses INET it will be more difficult and
the device driver will be required. The AT&T Blit(a programmable graphics
terminal) used a device driver approach to solving this problem in Sys 5.2, 
the driver was called mpx, see the BLTJ Oct 84 for more details on this.

I guess the bad news is that it is unlikely you will see DNET on Sys 5.2,
although I agree it would be nice. On the other hand DNET should crop
up on Sys 5.3 and 5.4 without too much effort. 

It would be a fun project, Ahh if only I didn't have to sleep!! 8-)


p.s.

This article was composed on my 2500 running AmigaUUCP .50, are their
many other users of this sofware out there? I have been using this
package to receive mail and news for several months now. My only problem
with it is the newsgroups fill up to quickly!! Is anyone working on 
an upgrade to the software?


--
      /\  George MacDonald, Pulsar Software Inc. - Unix Software Consultants
_____/  \    _____  ..{harvard,decvax}!zinn!pulsar!gmd or gmd@pulsar.mv.com
         \  /       
          \/   36 Hickory Ln,  Windham, N.H.  03087  Tel (603)-434-2614

shadow@pawl.rpi.edu (Deven T. Corzine) (08/13/89)

In article <907@corpane.UUCP> sparks@corpane.UUCP (John Sparks) writes:

John> Many Amiga/Unix users will appreciate it if DNET on SYS V is
John> possible.  Does it/will it exist?

On 11 Aug 89 04:06:49 GMT, gmd@pulsar.UUCP (George MacDonald) said:

George> Porting DNET to anything before 5.3 without the socket library
George> will be a chore. The major problem in emulating the socket
George> functions will be providing multiplexed input on a single
George> channel(serial line).  This needs to be done in the kernel to
George> do it properly, hence a device driver is needed.

SysV has nothing providing the functionality of select()?

George> I am not familiar with DNET internals, and not sure which
George> address domain it uses (UDP, INET ...) or which socket
George> communication style(stream or datagram).

Unix-domain sockets.  Named pipes, essentially.

George> If DNET uses UDP and stream sockets then emulation using named
George> pipes seems at least possible.

The obvious choice; Unix domain sockets seem virtually identical to
named pipes, except for the interface, probably.

George> If DNET uses INET it will be more difficult and the device
George> driver will be required.

I'd rather DNet use INET than clutter my filespace.  *sigh*

George> I guess the bad news is that it is unlikely you will see DNET
George> on Sys 5.2, although I agree it would be nice.

If you can replace select() with something and used named pipes, it
should be perfectly possible.

George> On the other hand DNET should crop up on Sys 5.3 and 5.4
George> without too much effort.

Well, on SysV.4 it should compile straight...

Deven
--
Deven T. Corzine        Internet:  deven@rpi.edu, shadow@pawl.rpi.edu
Snail:  2214 12th Street, Troy, NY 12180       Phone:  (518) 271-0750
Bitnet:  deven@rpitsmts, userfxb6@rpitsmts     UUCP:  uunet!rpi!deven
Simple things should be simple and complex things should be possible.

thad@cup.portal.com (Thad P Floryan) (08/14/89)

Re: BSD sockets on SVR2 and &tc, the following extract was JUST posted to
the listed newsgroups, and the postings contain everything needed to bring
up sockets (and ptys and vtem and &tc) on the UNIXPC (aka 3B1).  This is a
source distribution, and it appears the Makefile requires one to use gcc.
Some of the sources have UCB (BSD) copyrights in their headers.  Seems if
someone wants to make DNET function with System V *NOW*, these sources would
be an excellent starting point.

Thad Floryan [ thad@cup.portal.com (OR) ..!sun!portal!cup.portal.com!thad ]

-------------------------included posting follows-------------------------

From: bes@holin.ATT.COM (Bradley Smith)
Newsgroups: comp.sys.att,unix-pc.general,unix-pc.sources
Subject: Announcement Message for Unix domain socket code & pty
Message-ID: <647@holin.ATT.COM>
Date: 12 Aug 89 16:33:50 GMT
Date-Received: 13 Aug 89 05:56:41 GMT
Organization: AT&T Bell Labs, OIL, Holmdel, NJ

This is an announcement message stating that in unix-pc.general
(and comp.sys.att for non unix-pc sites) that the socket/pty code
will be posted.  It is in 5 shar files, and below is the ls -ls
of these files.

 132 -rw-r--r--   1 bes      HSJ        66712 Aug 12 12:29 uipc.p1
  88 -rw-r--r--   1 bes      HSJ        44418 Aug 12 12:23 uipc.p2
  81 -rw-r--r--   1 bes      HSJ        40654 Aug 12 12:24 uipc.p3
  64 -rw-r--r--   1 bes      HSJ        32097 Aug 12 12:25 uipc.p4
  94 -rw-r--r--   1 bes      HSJ        47540 Aug 12 12:29 uipc.p5

If you don't getit send me mail.

Have fun, brad
-- 
Bradley Smith
Computer Systems Offer Integration Laboratory
AT&T Bell Labs, Holmdel, NJ 
201-949-0090 att!holin!bes or bes@holin.ATT.COM