draft-ietf-pilc-link-09 - comments.
Gorry Fairhurst <gorry@erg.abdn.ac.uk> Fri, 15 March 2002 16:50 UTC
Message-ID: <3C92266C.7A156AA5@erg.abdn.ac.uk>
Date: Fri, 15 Mar 2002 16:50:51 +0000
From: Gorry Fairhurst <gorry@erg.abdn.ac.uk>
Organization: ERG
X-Mailer: Mozilla 4.73C-CCK-MCD {C-UDP; EBM-APPLE} (Macintosh; U; PPC)
X-Accept-Language: en
MIME-Version: 1.0
To: "pilc@grc.nasa.gov" <pilc@grc.nasa.gov>
Subject: draft-ietf-pilc-link-09 - comments.
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
X-ERG-MailScanner: Found to be clean
Sender: owner-pilc@grc.nasa.gov
Precedence: bulk
Status: RO
Content-Length: 6111
Lines: 238
Please see comments below on Link -09. I find the text mostly correct and a significant improvement on previous drafts. Comments are arranged in section order. Best wishes, Gorry Fairhurst ------------ OVERALL TERMINOLOGY! It may be good to replace "packet" with "IP packet" unless you specifically refer to IPv4 or IPv6 - since some subnetworks also call the SNDU a "packet" ;-). BE CAREFUL though in the section on "buffering" since packet refers to either of the two cases!!! "Router" or "Gateway"?????? - Both terms are used. I would suggest "router"? ----------- INTRODUCTION ENGLISH Introduction para 5, after list of 3 points - you start a sentence with "But", consider rephrasing to remove this. QUERY: "rely" or "benefit significantly from"? In the sentence which follows - does IP QoS and mcast "rely" on these or may it just "benefit" from support? - I'd argue you can multicast with no support, it's just inefficient - and QoS can be done solely with queues - it's just better if the subnet also has a notion of QoS. If this is really so, you may want to add these comments to the later para starting "However". QUERY TERM: "subnetwork?" We have been engaged in a deep and confusing discussion with an IESG member - part of which turned out to originate in loose definitions of terms. I think we *should* define "subnetwork" explicitly in the introduction to avoid such problems. ------------ MTU QUERY TERMS: "Router fragmentation" and "internal fragmentation" para 6 I don't think the term "router fragmentation" is clear. In the ARQ draft, we say "Explicit network layer IP fragmentation" instead of "router fragmentation" - to make sure we do expect transparent link fragmentation to still occur in some subnetworks. Here's our current text to compare with the link draft "internal fragmentation": "To allow for efficient use of the channel, the maximum link frame size may therefore be considerably lower than the maximum IP datagram size expressed in the IP Maximum Transmission Unit (MTU). Each frame will then contain only a fraction of an IP packet and transparent implicit fragmentation of the IP datagram is used [DRAFTKARN01]. A smaller frame size introduces more frame header overhead per payload byte transported." - In this ID, it may be better to use "subnetwork frame". - I haven't come across the term "internal fragmentation" - I guess "Link" or "transparent" fragmentation are more common terms. Is "subnetwork fragmentation" the best term for this ID? Last para of section. "IP fragmentation" --> should be the same terms as you finally choose for "router fragmentation". ------------ Choosing the MTU para 6 starting "A third example..." "internal fragmentation" -> "subnetwork fragmentation" (see above). ------------ Framing Para starting "in AAL5.." This para should be joined to previous. Word "cel" =-> "cell" ------------ Connection-oriented Para 1 - Missing reference for TCP [RFC793]. Para 2 - Are these subnetwork "nodes" or "IP routers"??? - I think the text is talking about the number of IP Routers in a subnetwork, not the number of internal nodes. ------------ Broadcast BEWARE of "point-to-point" and "shared" - that was one of the terms that caused a lot of questions in ARQ. This is *NOT* the same as a point-to-point PPP connection! We now use the term "dedicated link" instead of "point-to-point". ------------ BOD para 1. "Spectrum resources" -> replace by "spectrum resources (i.e., the shared channel that the radio link uses)" ---------- TCP v LINK... ENGLISH" Style becomes very tutorial in this section: Para 2. "For largely historical reasons" - doesn't seem to add anything - delete? "ARQ has been used for many decades..." - delete? "Unlike FEC..." - delete? - replace next "It" by "ARQ". "FEC is also..." - delete? - This section is still probably too long, given we have another ID on the subject - but *I'm* not doing the pruning... --------- Recovery NOTE: IN ARQ we now define "outage" as " outage (i.e. a period during which the link loses all or virtually all frames, until the link is restored)". Suggest you replace "link" by "subnetwork" and insert this at the start. ENGLISH "And" Para 3, sentence starts with "And" - could be rephrased. --------- How TCP works NOTE: This section still reads strange since it come in the middle of the document, well after we have talked about many TCP issues!!! para starting "TCP Assumes" Combine paras 2 and 3 (next). para 9: " A new technique" This has the use of "MUST" etc in the paragraph - no other paragraphs in this ID use this style. - suggest you replace with normal type. ----------- Analysis NOTE: The example seems awkward appearing at the mid-point of the document. Observation 2: "link-layer" -> "subnetwork" "bit error rate" -> "bit error ratio" --- this is WRONG, it is NOT a rate! "PACKET_SIZE" really is a very ***BAD*** term (I said this before). Since we can and often **DO** have transparent link fragmentation, this is not a good way to talk about things in general. You need to EXPLICITLY say that you "consider only the case where one packet is ent per subnetwork frame (i.e. no transparent link fragmentation)". You **SHOULD*** also change PACKET_SIZE to FRAME_SIZE - since this is clearly not just an IP packet!!! This current text is very misleading and seems at first sight to contradict other parts of the document - fixing the terminology will make the real issue clear. --------- QoS para 9 is just one sentence long. - Combine with previous? para 9 : ToS isn't defined - it should be: "Type of Service (ToS) [RFC791]" --------- Delay para 5 is just one sentence long. - Combine with previous? para 5: Starts with "But" - delete the "But". para 11 is just one sentence long. - Combine with previous? --------- Bandwidth Asym para 1, line 6 insert "," after acknowledgements" near end of the line. --------- Buffering para 2 "a complete IP datagram or a subnetwork packet" - terminology problem, suggest: "a complete IP packet, or a subnetwork frame" ----------
- draft-ietf-pilc-link-09 - comments. Gorry Fairhurst