Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!rutgers!tut.cis.ohio-state.edu!ucbvax!VAX.FTP.COM!jbvb
From: jbvb@VAX.FTP.COM (James Van Bokkelen)
Newsgroups: comp.protocols.tcp-ip.ibmpc
Subject: Re: Wide Area Network
Message-ID: <8906201335.AA10983@vax.ftp.com>
Date: 20 Jun 89 13:35:40 GMT
References: <1300@lafcol.UUCP>
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 31


   Date: 19 Jun 89 17:43:14 GMT
   From: Patrick Ciriello II <usc!polyslo!vlsi3b15!lehi3b15!lafcol!ciriello@apple.com>
   ...
   o	Bridging Token Rings to 802.3 LANs

You can't.  Period.  No way.  First of all, the bridge would need to
handle RIF fields on the 802.5 side, somehow stuffing them into each
packet from the 802.3 side.  Now, given a sufficiently energetic bridge
manufacturer, and a sufficiently small 802.5 LAN, you could build a bridge
which found out who was where, but I don't think it would be loveable.
Second (and fatal), 802.5 is an MSB LAN, and 802.3 is an LSB LAN.  On
MSB LANs a MAC-layer address is sent in one bit-order in the header, and
the opposite bit-order when sent as data.  On LSB LANs, the bit-order is
the same in header and data.  Thus, while a bridge can fix up the bit-order
of the MAC addresses easy as pie, it can't know which data may be a MAC
address (e.g. in ARP packets, or network managment systems), so it can't
treat it specially.  End of story, thanks IBM.  See a paper by Roger Pfister
of BICC Data Networks entitled "Bit Ordering in MSB and LSB LANs", presented
to the ISO FDDI working group last year, for more detail (I can post it if
people want).

   o	TCP/IP in a Token Ring environment

Works fine.  IBM does it on AIX, DOS and mainframes, Proteon, cisco and
IBM sell routers with 802.5 interfaces, we sell a DOS 802.5 TCP/IP, Sun
has promised 16Mb ring support.  See RFC 1042 for how to encapsulate IP.
   ...

James B. VanBokkelen		26 Princess St., Wakefield, MA  01880
FTP Software Inc.		voice: (617) 246-0900  fax: (617) 246-0901