Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!sdd.hp.com!hp-col!hpctdlb!chw
From: chw@hpctdlbcol.hp.com (Charlie Whiteside)
Newsgroups: comp.protocols.tcp-ip
Subject: Re: ethernet alignment errors
Message-ID: <3270026@hpctdlbcol.hp.com>
Date: 30 May 91 18:45:39 GMT
References: <1991May30.022715.22240@granite.ma30.bull.com>
Organization: Hewlett-Packard CTD, Colo. Spgs.
Lines: 27

/ hpctdlb:comp.protocols.tcp-ip / piacenti@granite.ma30.bull.com (Paul Piacentini) /  8:27 pm  May 29, 1991 /
>  Most of our ethernet equipment, analyzers, etc report counters on bytes,
>packets, CRC errors, collisions, runts, giants, and frame alignment errors.
>I'm a little fuzzy on exactly what consistutes a frame alignment error.
>Is it a packet that was transmitted within the minimum IEEE inter-packet delay
>spec window (9.6 usec?) ?

No. 802.3/Ethernet frames are required to send their data in 8 bit units called
octets or bytes. If the frame ends (CSN goes false) and the number of bits is
not divisible by eight then the frame had an alignment error.  An alignment
error can be divided into two categories. (These are terms we use here, the
industry may use variations on these) a. Dribble frame - This occurs when a
frame is not octet aligned but the FCS was good. Usually means the CSN line
went false a few bits after the frame ended, FCS good shows that data integ-
rity for the frame exists.  b. Misaligned frame - Non-octet aligned frame 
with bad FCS. Could be an extra bit(s) was added in the data field, caused
shift of data, FCS caught the extra bits.

Inter Frame Spacings of less than 9.6 uS is a totally different measurement.

Hope this helps.

Charlie


chw@hpctdlb.hp.com
719-531-4388