Re: re the plenary discussion on partial checksums
> Why, oh WHY would I want to receive a known corrupted packet ? why oh why would you ever want to talk with someone over a phone that occasionally clicked or popped? why or why would you ever want to watch a video with snow, or an occasional missing pixel, or even an occasional missing frame? sometimes getting a lossy packet is better than not getting one at all, or having to wait for a retransmission.
Re: re the plenary discussion on partial checksums
> >so it seems like what we need is a bit in the IP header to indicate > >that L2 integrity checks are optional, and to specify for various > >kinds of IP-over-FOO how to implement that bit in FOO. > > > How would an app know to set this bit? The problem is that different > L2s will have different likelihoods of corruption; you may decide that > it's safe to set the bit on Ethernet, but not on 802.11*. And, in > general, the app doesn't know all of the L2s that may be involved when > it sends a packet. I'm not sure that the app needs to know how much lossage to expect, or to specify how much lossage is too much. It just wants the bits, errors included. Depending on the app's needs it might dynamically adapt to varying degrees of error by adding its own FEC, e2e retransmission, and/or interleaving, and this probably works better than trying to have the app either determine statically how much error it can expect or by having the app specify to the network how much error is acceptable. I suppose we could define a maximum error rate (say 15%) that IP-over-FOO should be designed to provide if the "lossage okay" bit is set. But practically speaking I doubt it's necessary to do that- links that are designed to support lossy traffic will already have enough FEC or whatever to suit that kind of traffic. The biggest questions I have are: - where to put this bit? - are there unintended consequences of doing this that we can forsee? Keith
Re: re the plenary discussion on partial checksums
> > so it seems like what we need is a bit in the IP header to indicate > > that L2 integrity checks are optional > > A lot of folks seem to forget that from the point of view of IP L2 > includes the busses between memory and the L2 network interface. > There have been more than a few recorded cases where packet errors > were introduced as the packet flowed in or out of memory, unprotected > by link CRCs. For apps that tolerate lossage (or more precisely, for apps that work better in the face of some transmission errors than they do if all packets with transmission errors are dropped), it doesn't matter whether the errors occur in the memory-to-interface link or somewhere else - they'll deal with the errors no matter where they occur. Of course, the apps that can't tolerate lossage won't set the "lossage is okay" bit, and they'll continue to expect that the packets that do arrive, arrive intact. For one particular L2 technology X this might simply mean that packets that don't have that bit set, but do have errors, are dropped. For another L2 technology Y it might mean that if that bit is not set then the IP-over-Y spec will require FEC or link-level retry or both to make sure that those packets have a reasonable probability of getting there intact. > To my way of thinking we don't need a bit in the IP header, we need a > bit in the heads of implementors to remind them that relying on > link-by-link protection can be dangerous even if the links have strong > CRCs. Actually, it seems like we need a bit in the heads of people who don't understand that - some kinds of links have inherently high error rates, - some apps are capable of dealing with less-than-perfect data, - adding FEC and/or link-level retry to get error rates down to the level we're accustomed to from wire or fiber carries with it a substantial penalty in bandwidth and/or delay - we'd like to be able to use those kinds of links with IP, - we'd like to be able to run those apps over IP, and over those links, without paying the bandwidth or delay penalty for apps that don't need it. - we'd like a stable spec for this so we can carve it in stone, (er, silicon) - since it's going to be carved in stone (silicon) we would do well to get it right. Yes, this is a change to IP, and to the IP architecture. But it's not rocket science, and it doesn't have to affect things that don't use it explicitly.
Re: re the plenary discussion on partial checksums
On Wed, 16 Jul 2003, Keith Moore wrote: > so it seems like what we need is a bit in the IP header to indicate that > L2 integrity checks are optional A lot of folks seem to forget that from the point of view of IP L2 includes the busses between memory and the L2 network interface. There have been more than a few recorded cases where packet errors were introduced as the packet flowed in or out of memory, unprotected by link CRCs. To my way of thinking we don't need a bit in the IP header, we need a bit in the heads of implementors to remind them that relying on link-by-link protection can be dangerous even if the links have strong CRCs. > ... IP option to provide a stronger checksum than normally exists The last time I saw a comparision of checksum algorithm strengths was back in the OSI days when the IP checksum was compared to the OSI Fletcher checksum (my memory is that the IP checksum came in second.) --karl--
Re: re the plenary discussion on partial checksums
Ok, I have to ask a silly question (not like that would be a first on this list) Why, oh WHY would I want to receive a known corrupted packet ? Are we talking about someone thinks they can eeke out 1% more performance because their phy/mac can cut over immediately rather than wait for the packet and verify the checksum ??? (or compute it on the sending side) I guess I don't see the benefit, I guess rather than a hardware L2 check, you rely on something in your hardware later up to fail a check (including a L7 protocol) and drop the frame there ??? I wish I had been there to see the discussion Bill On Wed, Jul 16, 2003 at 04:21:47PM -0400, John Stracke wrote: > Keith Moore wrote: > > >so it seems like what we need is a bit in the IP header to indicate that > >L2 integrity checks are optional, and to specify for various kinds of > >IP-over-FOO how to implement that bit in FOO. > > > > > How would an app know to set this bit? The problem is that different L2s > will have different likelihoods of corruption; you may decide that it's > safe to set the bit on Ethernet, but not on 802.11*. And, in general, > the app doesn't know all of the L2s that may be involved when it sends a > packet. > > -- > /==\ > |John Stracke |[EMAIL PROTECTED] | > |Principal Engineer|http://www.centive.com | > |Centive |My opinions are my own.| > |==| > |Linux: the Unix defragmentation tool. | > \==/ > >
Re: re the plenary discussion on partial checksums
Keith Moore wrote: so it seems like what we need is a bit in the IP header to indicate that L2 integrity checks are optional, and to specify for various kinds of IP-over-FOO how to implement that bit in FOO. How would an app know to set this bit? The problem is that different L2s will have different likelihoods of corruption; you may decide that it's safe to set the bit on Ethernet, but not on 802.11*. And, in general, the app doesn't know all of the L2s that may be involved when it sends a packet. -- /==\ |John Stracke |[EMAIL PROTECTED] | |Principal Engineer|http://www.centive.com | |Centive |My opinions are my own.| |==| |Linux: the Unix defragmentation tool. | \==/
Response from a former IMPP Chair (Re: Last Call: A Model forPresence and Instant Messaging to Proposed Standard)
WRITTEN IN MY ROLE AS FORMER IMPP CHAIR Dave and Marshall, I have issues with your presentation of reality. I am unable to see technical issues of real substance in your comments; the issues seem all to be procedural, and revolve around the missing update to the WG charter. Since I was one of the WG chairs at the two IETF meetings immediately following the "group of nine" effort, I accept responsibility for this procedural error. The charter should have been updated. However, I must take issue with some of your presentation of how the WG decided issues, as well as some of your specific issues. In particular, your statement: The problem with the group's dichotomy between end-to-end vs. gateway goals was discussed repeatedly. Only towards the end of the working group's effort did this appear to become resolved -- in the direction of an end-to-end content standard. However, even then it was clear that the few remaining participants in the group continued to hold very different understandings of the goal. is directly contradicting the minutes of the December 15, 2000 meeting of the WG in San Diego, which say: ROOT: Should messages passed wround in CPIM-compliant Q: do we think that we need to be able to pass around a pile of bits that can be signed: A: Rough consensus in the room (all but 1 raised hand yes) Q: Should CPIM specify structure of this message? A: All think it should specify the format. The exact format was thereafter discussed in March 2001 and August 2001. Because of (I believe) the failure of the editor to update the core CPIM document between February 2001 and November 2001, the group did not meet in December 2001; the format was discussed again in March 2002, July 2002, November 2002 and November 2002. At no time do the minutes record a decision of the group to revisit or reverse its decision to support a single format for interoperability. For a further instance, consider this complaint: How can "duration" have any useful meaning when there is no baseline reference for the starting point or ending point of the duration and when Internet exchange latencies are completely unpredictable? In other words, when a participant receives a duration value from another participant, what does it mean? Duration relative to what point of time? We do not know how many seconds it took for the service data to reach the receiving participant. The decision to use intervals was a very visible one in the WG, and none of the minutes I have read show this decision being challenged at any working group meeting; indeed, instead we see long wrangles over the meaning of "duration = 0", which is hard indeed to express in a format that has a "baseline reference". While I have not found this in the minutes, I believe the justification for using duration was the same as that used for the sysUptime in SNMP: That one should NOT require the elements of an IM system to have synchronized clocks. A common thread running through the minutes of IMPP meetings is the perception that all major issues have been settled: San Diego, December 1999: The goal is that the group will finish its work and then go away, hopefully before the next IETF. London August 2001: LD: We've reached some kind of closure on all major points; really hope this will have been the last IMPP meeting (i.e., that we can wrap this into revised & finished documents before the next IETF). My conclusions: The working group has suffered from very slow document updates, a bad error in judgment (mine) re charter update, and repeated re-raising of old closed issues (for instance, at Atlanta in November 2002, Dave Crocker could be heard re-raising the issue of the need for loop control, which the group had discussed and decided in December 2000, choosing hopcount as the preferred mechanism in March 2001). However, I find the criticisms raised against the process leading to the forwarding of these documents to the IESG to be very much off target. Harald Alvestrand Speaking as a former chair of the IMPP group
re the plenary discussion on partial checksums
I'm not sure what the problem is here: - UDP checksums are optional - optional checksums probably aren't applicable to TCP - IPv4 has IP header checksums - IPv6 could define an option for IP header checksum (could be applicable to IPv4 also, if you want a stronger checksum for the header) - whether L2 drops packets whose checksum fails is an L2 matter; the ability to turn this on or off is an L2 feature - if the app needs a different integrity check than all-or-nothing, this is something that belongs in the app protocol so it seems like what we need is a bit in the IP header to indicate that L2 integrity checks are optional, and to specify for various kinds of IP-over-FOO how to implement that bit in FOO. and maybe that bit could go in the IP option to provide a stronger checksum than normally exists in the IP header (so that the header, at least, is protected) Keith
Re: Multicast Last Mile BOF report
let's get real here. though we have been pushing it since i was in nappies (and i have deployed it in isp(s), and i also donate to hopeless progressive causes), there is far more email traffic about multicast than there is actual multicast traffic on the WAN internet (yes, it is heavily used on a few LANs). randy
Re: Multicast Last Mile BOF report
Keith Moore wrote: maybe the ISPs supporting multicast could prioritize that traffic, thus providing better service on multicast than unicast, thus providing an incentive for receivers to use multicast over unicast. This would provide an incentive to game the system and use multicast even if you had only one receiver. Isn't the real problem the fact that the set of streams that people might receive is so diverse, most links won't be carrying more than one copy of the unicast stream anyway? Multicast would help only at the sender's end, in the case of flash crowds; and the sender's ISP has no incentive to shrink the size of the bitpipe the sender needs to buy. -- /==\ |John Stracke |[EMAIL PROTECTED] | |Principal Engineer|http://www.centive.com | |Centive |My opinions are my own.| |==| |"But she calls her ship _Mercy of the Goddess_!" "Kali." "Oh."| \==/
Re: Multicast Last Mile BOF report
] >> and 500 gadgets do not make a technology adopted when it has no ] >> business model folk can understand. ] > one business model that might be understandable is: you should support ] > multicast if/when it saves you enough bandwidth (over the same content ] > being sent over separate unicast streams) to make it worth your cost. ] ] if i needed to save bandwidth, which is not demonstrated as i sell it ] , if your customers pay flat rate and you can deliver them the same or better service at less cost to you, isn't this at least some incentive? (perhaps not enough to overcome your costs?) if your customers pay per megabyte and they can get better service at less cost by using multicast, isn't this also an incentive of a different sort? ] that there is no content my customers want means i will add ] capex and opex to save no bandwidth. indeed, lack of content is a problem. multicast won't fly unless/until there is content available through it that lots of people want to get. IETF meetings are great for me but they have a limited audience overall.
Re: Multicast Last Mile BOF report
] Users will favour those ISPs which prioritize unicast and pay more money. more money will nearly always buy more bandwidth. ] Other users may use the prioritized multicast for 1:1 communication. the trick is to only prioritize multicast for which there are enough listeners. but this seems doable.
Re: Multicast Last Mile BOF report
Keith; > ] > one business model that might be understandable is: you should support > ] > multicast if/when it saves you enough bandwidth (over the same content > ] > being sent over separate unicast streams) to make it worth your cost. > ] > ] It means it saves nothing for receivers that they don't want to > ] receive multicast stream. > ] > ] Then, senders are not motvated to send multicast stream. > > maybe the ISPs supporting multicast could prioritize that traffic, thus > providing better service on multicast than unicast, thus providing an > incentive for receivers to use multicast over unicast. Prioritization is orthogonal to uni/mulitcast issue. Users will favour those ISPs which prioritize unicast and pay more money. Other users may use the prioritized multicast for 1:1 communication. Masataka Ohta
Re: Multicast Last Mile BOF report
On Wed, 16 Jul 2003 22:48:06 +0859 () Masataka Ohta <[EMAIL PROTECTED]> wrote: ] > one business model that might be understandable is: you should support ] > multicast if/when it saves you enough bandwidth (over the same content ] > being sent over separate unicast streams) to make it worth your cost. ] ] It means it saves nothing for receivers that they don't want to ] receive multicast stream. ] ] Then, senders are not motvated to send multicast stream. maybe the ISPs supporting multicast could prioritize that traffic, thus providing better service on multicast than unicast, thus providing an incentive for receivers to use multicast over unicast.
Re: Multicast Last Mile BOF report
>> and 500 gadgets do not make a technology adopted when it has no >> business model folk can understand. > one business model that might be understandable is: you should support > multicast if/when it saves you enough bandwidth (over the same content > being sent over separate unicast streams) to make it worth your cost. if i needed to save bandwidth, which is not demonstrated as i sell it , that there is no content my customers want means i will add capex and opex to save no bandwidth. randy
Re: Multicast Last Mile BOF report
Keith; > ] and 500 gadgets do not make a technology adopted when it has no > ] business model folk can understand. > > one business model that might be understandable is: you should support > multicast if/when it saves you enough bandwidth (over the same content > being sent over separate unicast streams) to make it worth your cost. It means it saves nothing for receivers that they don't want to receive multicast stream. Then, senders are not motvated to send multicast stream. Also, the bandwidth saving is beneficial to ISPs that ISPs supporting multicast should charge less to its customers. It's not bad if ISPs are competitive and backbone bandwidth were expensive. Masataka Ohta
Re: Multicast Last Mile BOF report
] and 500 gadgets do not make a technology adopted when it has no ] business model folk can understand. one business model that might be understandable is: you should support multicast if/when it saves you enough bandwidth (over the same content being sent over separate unicast streams) to make it worth your cost.
Re: The requirements cycle
Alex, [clipped]... > Again, I was not on the IESG when PPVPN was started, however, I > think that naturally well-scoped technologies with a clear direction > in the solution space very often do not need requirements and > frameworks. > It seems that the VPN problem and solution spaces are large and > complex enough to warrant both requirements and framework documents. > That said, these documents do not have to be long and fat, and it > should be possible to produce an acceptable quality document within > 6 months. > Regarding IESG feedback (where my piece was probably the biggest): > Predicated on the assumption that reqs/fw documents are not needed, > any feedback, whether it is from the IESG or not, will be perceived > as a rock fetch. If we assume those are useful, IESG review is part > of the process of ensuring high quality of these documents. In fact the reason why the PPVPN wg was created is because there were two solutions (VR and 2547) already defined with some known *deployment* already happening. That's what motivated the ADs to scope the WG to just standardize these two solutions which at that time the WG was called "NBVPN" for network-based VPNs. There was in fact no need for the framework and requirements drafts since day one there was no objective to create the "best/optimum" approach and the wg didn't debate at all the question of which one is best VR versus 2547. It was pretty much left to the market to decide. I am pretty sure that the feedback on that was given initially to the ADs/IESG, and I don't understand why it was ignored (at least from my perception). I think the chairs were just following the advice of the IESG on that (that the framework and requirements are necessary before any solution is considered). If you notice the ppvpn charter included a statement that indicates that no new protocols will be developed. This was added because of the existence of the solutions (well before the working group was created), and there was a feeling that if the wg allows for new protocols, etc, the delay for getting the solutions standardized for the providers would have been much bigger. It is ironic that the wg members initially tried to address the potential technical reasons that can happen for delaying the work but couldn't predict that the IESG request to develop the framework and requirements drafts are mostly the reasons for the actual delay. And it is unfortunate that the recent ppvpn decisions just ignored and didn't explicitly acknowledged that fact. I think the delay situation and its impact on the IESG decision on what to do with ppvpn wg should have taken into account the actual history of ppvpn working group... Hamid.
Re: no multicast of IETF plenaries
Basically we were unable to use the same rooms for the plenary the we use for the regular sessions. For us it's just to much to move the ch1 equipment in two hours. It is possible the the equipment I,m using will be able to source mpeg-1 but not the other sources from the plenary . if it works I'll update the webpage at least two hours before the plenary. joelja On Tue, 15 Jul 2003, Keith Moore wrote: > from http://videolab.uoregon.edu/events/ietf/ietf57.html: > > > NEWS: > > > > 07-15-2003 - We're not going to broadcast the two plenary > > sessions in the evening wednesday and thursday. They are in > > different rooms for our equipment, and moving it is too hard. > > Currently we plan on recording the two plenaries and they should > > be accessible later. > > I'm sorry to learn this. The multicast access is actually turning out > to be quite useful for me this time around, and the plenaries are > important parts of the meeting. I hate to miss them. Watching the > recorded meetings isn't the same because there's no way to provide > real-time feedback. > > (fwiw, in my experience Jabber isn't a very good mechanism for keeping > track of what's going on, but it has worked quite well for making the > occasional remotely-originated comment.) > > despite the inability to multicast the plenaries, I'm very appreciative > to the multicast team for allowing me to participate in some of the WG > meetings from Tennessee. > > Keith > > -- -- Joel Jaeggli Academic User Services [EMAIL PROTECTED] --PGP Key Fingerprint: 1DE9 8FCA 51FB 4195 B42A 9C32 A30D 121E -- In Dr. Johnson's famous dictionary patriotism is defined as the last resort of the scoundrel. With all due respect to an enlightened but inferior lexicographer I beg to submit that it is the first. -- Ambrose Bierce, "The Devil's Dictionary"
