[RPRWG] Comments on Definitions
The following are my comments against the file bs_terms_03a.pdf, posted on
the web site by Bob Sultan. I probably won't have a chance to be at the
Monday 9 - 11am meeting. Please review these comments there.
R.D. Love's comments against file bs_terms_03a.pdf
network (data transmission points interconnected by communication channels.
Proposed change: "points" to "nodes" in definition #2.
Resilient Packet Ring definition: Footnote question: Should this be
specifically the protocol standardized by IEEE802?
My answer, YES! When used in our standard, it ought to refer specifically to
standard implementations.
station (data station): [adapted15 from IEEE 100 (C/EMB/LM) 1073.3.1-1994,
1073.4.1-1994, 8802-5-1995] A device that may be attached to a network for
the purpose of transmitting and receiving information on that network. A
station is identified by a MAC address.
Change the last sentence to:
A station is identified by one or more MAC address. - See the definition of
dual ring station for rationale behind this request. I believe that some
dual ring stations may be built with at least two MACs (as was done in IEEE
802.5.)
Encoding (The following definition appears in 802.5: 'symbol: [802.5-1998
1.3.68] In this standard a symbol consists of two signal elements. Four
symbols are defined: data_zero, data_one, non-data_J, and non-data_K.'. Does
this (or some analog of this) apply to 802.17?) NO!
and
signal element:: [802.5-1998 1.3.61] The logical signal during one half of a
bit time which may take on the values of Logic_1 or Logic_0.
These are definitions that are 802.5 specific. Since we will be picking up
existing PHY layers, the encoding and the characteristics of the symbol
elements will be variable. Delete the definition of signal element until we
see that we need one.
layer agnosticism - Mark these terms as ones to be deleted when the draft
goes out to ballot. They are useful in understanding our objectives.
link: (1) [IEEE 100 (C/LM) 802.5c-1991] A unidirectional physical and media
connection between two stations. (2) [IEEE 802.3-2000 1.4.153] The
transmission path between any two interfaces of generic cabling.
Based on the definitions of link segment and link partner, the second
definition of link should be removed.
Footnote 31: A link aggregation group appears to the MAC Client as a single
link, so I assume that a link can refer to a link aggregation group. Is that
correct? - Based on definition 1, the assumption appears correct.
ring: (1)The collection of stations and links forming a resilient packet
ring. (2) A set of congruent ringlets forming a resilient packet ring
- NO! You have now defined a ring as a MAC protocol (refer to the
definition of resilient packet ring).
- I believe you need a definition that more specifically talks about the
allowable topology, i.e.
I would suggest defining a ring as a set of congruent ringlets, at least two
of which are opposing. I see that this is what your definition is for
multi-ring. I would suggest dropping the definition of multi ring.
ring medium: The abstraction of a ring as a continuous transmission medium
having a closed path.
Or more simply: The transmission medium used by a ringlet or congruent
ringlets.
ring segment (segment) : (1) The portion of a ring bounded by two stations
interconnected by one or more links. (2) [802.5-1998 1.3.57] Section of
transmission path bounded by repeaters or converters38. Ring segment
boundaries are critical for determining the transmission limits that apply to
the devices within the segment. (3) [IEEE 100 (C/LM) 802.3u-1995,
802.9a-1995] The portion of a ringlet39 between the producer and consumer
along which a packet is sent. The segment traversed by a send packet is the
send segment, and the segment traversed by an echo is the echo segment. (4)
[IEEE 100 (C) 610.6-1991] The medium connection40, including connectors,
between medium dependent interfaces in a LAN.
Drop Definitions 2 and 3.
dual-ring: (1) [(C/LM) 802.5c-1991] A topology in which stations are linked
by link pairs and, from any one station to another, there are exactly two
distinct paths, where a path is defined as a sequence of consecutive links in
which no link pair is traversed more than once. (2) A ring composed of
exactly two congruent ringlets whose transmission paths have opposite
orientations.
Drop Definition 1. Definition 2 for dual ring is much better.
<dual-ring station: [(C/LM) 802.5c-1991] A station that attaches link pairs
that have opposite directions of data flow with respect to the adjacent dual
ring stations. It consists of MACs, a crosspoint function, and dual ring
management. It may have other attachments.>
With this definition you run into trouble defining "station" as
having only a single MAC.
multi-ring: A ring composed of multiple congruent ringlets, at least two of
which are opposing ringlets.
I suggest dropping this definition in favor of ringlet and ring.
<upstream neighbor's address (UNA):: [802.5-1998 1.3.77] The address of the
station
functioning upstream from a specific station.>
Suggest adding at the end: "for a particular data flow"
broadcast address: [ISO/IEC2382-25 25.01.13] A group address that identifies
the set of all stations on the network.
Suggest changing "identifies" to "is recognized by" since a group
address has no knowledge of, and does not identify any stations on the
network.
Make the same change in "Multicast Address" though not in "individual
address".
For "data frame": define PAD.
maximum transfer unit (MTU): [IEEE 100 610.7-1995] The largest amount of data
that can be transferred across a given physical network.
Add "in a single frame" at the end of this definition.
data-stream: [(C) 610.10-1994] A continuous stream (sic) of data elements
being transmitted, or intended for transmission.
To avoid circularity, make that "A continuous field …
traffic: (1) The aggregate of frames circulating on a ring.
I think we need to add "in a specified time interval" for this definition
to really stand up.
<ring in : [802.5-1998 1.3.53] A port that receives signals from the main
ring path on the
trunk cable and transmits signals to the backup path on the trunk cable, and
provides connectivity to the immediate upstream ring out port.>
<ring out : [802.5-1998 1.3.55] A port that transmits the output signals to
the main ring path on the trunk cable and receives from the backup ring path
on the trunk cable, and provides connectivity to the immediate downstream
ring in port.>
The sins of my past are catching up with me. I believe I created these
definitions of ring in and ring out. I don't think that they are appropriate
for 802.17 and I would suggest we drop them entirely. Token ring had to
distinguish between trunk and lobe calbing, and between the primary ring path
and the backup path. Those concepts just don't apply to RPR.
strip (stripping) Definition 1 suffices.
definitions 2 and 3 are too limiting. If we don't use idles, then definition
4 is wrong.
change
copy (copying) : The receipt and acceptance of a frame from the ring by a
station after
determination that the frame is addressed to the station.
To
copy (copying) : The receipt and acceptance of a frame from the ring by a
station
The last part of the sentence (a) didn't make sense, and (b) would probably
be superfluous even if written well.
transmit (transmission)
Eliminate definitions 3 and 4.
encapsulation:
Change "must" to "may" in the definition, or delete the term "must"
altogether.
bit rate:
Delete definition 1. In an attempt to avoid a circular definition you use
the term "speed" which is properly a measure of distance/time, and not # of
bits/time.
bit time: (BT):
Leave out the example for bit time. We don't need an 802.3 example in an
802.17 definition list.
bandwidth: [FRF Glossary (modified)] The range of frequencies, expressed in
Kilobits per second, that can pass over a given data transmission channel
within a network. The bandwidth determines the rate at which information can
be sent through a channel - the greater the bandwidth, the more information
that can be sent in a given amount of time 62.
This definition is technically incorrect. Bandwidth is measured in Hz,
not in bits/second. Consider changing to a more accurate:
bandwidth: The range of frequencies, expressed in Hertz, used to pass data
over a given data transmission channel within a network. The bandwidth
together with the data encoding scheme determine the rate at which
information can be sent through a channel - the greater the bandwidth, the
more information that can be sent in a given amount of time.
Consider changing to:
time-to-live (TTL): Value carried in the protocol header of a PDU. It is
used to prevent a PDU that has not been stripped from the ring by its
destination or source station to be removed. The TTL value is generally set
to an initial value at the source and decremented at each subsequent hop. The
PDU is discarded when the TTL value reaches zero.
plug-and-play : Consider: The requirement that a station be capable of
transferring and receiving frames, and if permitted, sending frames without
manual intervention other than the physical connection of the equipment.
(With this definition I am trying to accommodate the conditions where the
station may or may not require permission to insert data frames onto the ring)
operational :
Delete definition 1. Definition 2 says it better and is complete.
Change
neighbor : [BH] A station that is exactly one segment away.
To: The station that is immediately upstream or downstream from the station
Or change "segment" to "link"
Because: Segment can include intermediate stations (see the definition we
have above)
partitioned ring
I would like to see a diagram. What do we call a ring with a break in it?
Is there a separate name?
path:
Use definition 1, and drop definition 2. I don't know how to use definition
2 for a multi-ringlet configuration.
topology database update (TDU):
Add the word Frame after update so we can differentiate the frame from the
information update.
topology discovery : The process by which the connectivity and capabilities
of the stations and links on the ring is discovered by a newly added station.
Delete "by a newly added station". Old stations must do topology discovery
every time there is a change to the ring.
insertion buffer [HvA-Alt] : A type of transit buffer used to store all or
part of a transit frame or frames awaiting completion of an ingress frame
transmission in progress.
Delete the two words "transit" from this definition. A transit buffer is
only defined with respect to transit frames. An insertion buffer holds the
data BEFORE it is placed on the ring, and therefore, the word "transit" is
incorrect.
ALSO, Remove the word "transit" from "
shaping buffer, for the same reason.
Replace both definitions of Latency with:
latency: The time it takes for information to get through a network, or
elements of a network, sometimes referred to as delay.
This definition combines the important features of the proposed definitions 1
and 2.
Drop the definition of
round trip propagation time:
frame check sequence (FCS) : is a field. Go with definition #3, except
change "remainder" to result. Both are correct, with remainder being more
correct, but a bit arcane for some readers.
flow control:
Use Definition 2 which does not presuppose how flow control will be done.
Backpressure: and throttle:
Change opposing ring, to opposing ringlet, to maintain consistency.
congestion management
Change "This" to "The set of procedures"
"This" does not refer to anything!
uncorrelated jitter:
Drop this definition. It has nothing to do with the jitter that we are
interested in for 802.17.
wander::
I wonder what the definition of wander is.
burstiness:
Change "Data" to "Stations"
Even better, burstiness is the uneven flow of data across a link or links, or
the uneven flow of data from a station.
For completeness change
delivered out-of-sequence frames [ITU I.)33.1 A.16 (modified)] A frame (Ft)
arriving at a destination station after a fame Ft+1, Ft+2, Ft+3…., Fn in a
sequence of frames F1, F2, F3, …. , Fn sent from a source station.
To
delivered out-of-sequence frames [ITU I.)33.1 A.16 (modified)] A frame (Ft)
arriving at a destination station after a fame Ft+1, Ft+2, Ft+3…., Fn, or
before fame Ft-1, Ft-2, Ft-3…., Ft-n, in a sequence of frames F1, F2, F3, ….
, Fn sent from a source station.
repeat [802.5-1998 1.3.50] The action of receiving a bit stream (for example,
frame, token, or fill) and placing it on the medium. Stations repeating the
bit stream may copy it into a buffer or modify control bits as appropriate.>
Delete the parenthetical "(for example, frame, token, or fill)" as this does
not apply to 802.17.
spatial reuse
The definitions are not specific and need help. Here is a general one for
spatial reuse.
spatial reuse A term referring to the capability of a ringlet to
simultaneously carry two or more data streams that are limited to
non-overlapping portions of the ringlet.
I need to better understand what global and local spatial reuse are.