home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.dcom.lans.ethernet:1808 comp.dcom.lans.misc:712
- Path: sparky!uunet!gatech!rutgers!news.cs.indiana.edu!ethome.et.iupui.edu!indyvax.iupui.edu!harvey
- From: harvey@indyvax.iupui.edu
- Newsgroups: comp.dcom.lans.ethernet,comp.dcom.lans.misc,comp.protocols.novell
- Subject: Re: 802.3 Non-Compliance (?)
- Message-ID: <1992Sep2.205749.1@indyvax.iupui.edu>
- Date: 3 Sep 92 01:57:49 GMT
- References: <1992Aug28.194558.4752@iplmail.orl.mmc.com> <1992Sep2.214854.28999@samba.oit.unc.edu>
- Sender: usenet@ethome.et.iupui.edu (USENET News System)
- Organization: Purdue University, Indianapolis
- Lines: 88
-
- In article <1992Sep2.214854.28999@samba.oit.unc.edu>, shava@hermes.oit.unc.edu (Shava Nerad Averett) writes:
- > In article <1992Aug28.194558.4752@iplmail.orl.mmc.com> more@tcc.orl.mmc.com () writes:
- >>In this meeting (which unfortunately I wasn't in), the vendor made the statement
- >>that Netware Servers violate the 802.3.
- >
- > This is the gods' own truth. Netware 802.3 (euphemistically called "802.3
- > raw" -- or "802.3 wrong" to some of us...;) has a frame type that looks
- > like this:
- >
- > Preamble 7 bytes
- > SFD 1
- > DestAddr 6
- > SourceAddr 6
- > Length 2
- > Data 46-1500
- > FCS 4
- >
- > Real 802.3, with 802.2, looks like this:
- >
- > Preamble 7 bytes
- > SFD 1
- > DestAddr 6
- > SourceAddr 6
- > Length 2
- > Data 46-1500
- > DSAP 1st byte
- > SSAP 2nd byte
- > CONTROL 3rd, 4th (opt)
- > Data and pad to minimum of 46...
- > FCS 4
- >
- > But the real problem is that Netware *always* sets the *length* field to
- > all F's, meaning that if you are (from what I understand) running DECNet
- > Phase V or OSI protocols which *believe* this field, you could end up
- > with a lot of confusion, because the packet will not be the length adver-
- > tised.
-
- No. The length field is filled in with the length OK. The problem is that
- where an Ethernet/802 controller expects to find the DSAP and SSAP fields
- are the first two bytes of the IPX header. This is the IPX checksum field.
- Since IPX checksums are never enabled, this field is always FFFF hex, which
- means "checksum disabled" to IPX. In the 802.2 standard SAPs of this form
- are reserved to the IEEE. This is what drives some Ethernet/802 controllers
- nuts. As far as VMS goes there is no way to even tell the Ethernet/802
- driver to receive these kinds of packets that I know of.
-
- > This is the only shred of verity I can find in the vendor's statement.
- > This has, however, nothing to do with collision detections and such.
- > Generally that's all in hardware...
-
- Yep...
-
- >>His statement was that when a netware
- >>server detects a collision, that rather than following the random backoff,
- >>in the Ethernet world, the Netware server immediately attempts to retransmit
- >>the collided packet(after sending the jam sequence). This vendor stated that
- >>because of this, TCP/IP based systems running on the same ethernet as Novell
- >>netware servers could experience some performance problems due to the
- >>noncompliant nature of the netware server. He further stated that for this
- >>reason, NASA has put a ban on Novell Netware at it's sites.
- >>
- > I don't know about NASA, but IPX and IP are the two major protocols on our
- > campus, and they don't seem to have any difficulty at all living on the
- > same wires. Could NASA be playing with OSI?
-
- No, I think the salesman just didn't know what he was talking about...
- Some large nets put restrictions on what protocols allowed on the backbone.
- Usually this is done simply to reduce routing overhead in the backbone routers.
- Perhaps that is what NASA did. Anyone from NASA here that knows?
-
- > Note, that Netware *can* use Ethernet_II rather than Ethernet_802.3 in
- > the configuration statement, and completely obviate even this argument.
- > Or Ethernet_SNAP...
-
- Yes. IPX and many other protocols including IP co-exist just fine on our
- network here at IUPUI. Many of our subnets were around long before they were
- connected to the backbone and have good reasons for staying with "raw 802.3"
- (e.g. they are using "raw 802.3" boot ROMs). All other subnets (including
- the backbone) use Ethernet II. Ethernet SNAP is not supported by our routers
- (Cisco AGS+s) for IPX (yet).
-
- IPX encapsulations, and getting IPX and IP to live together are frequent
- topics of discussion in the comp.sys.novell newsgroup.
-
- >[rest deleted]
- --
- James Harvey IUPUI OIT Technical Support
- harvey@iupui.edu uucp: iugate!harvey bitnet: harvey@indyvax
-