Talk:Network Time Protocol/Archive 1

What does this mran ?!
"...the amount of time it takes a proton to pass an electron"

How can one use NTP in Javascript ??. —Preceding unsigned comment added by Mac (talk • contribs) 19:24, 10 February 2003

Link to list of publich ntp servers
Hello,

I just came to this entry because I was searching for a list of public accesable time servers. Something like:

http://ntp.isc.org/bin/view/Servers/WebHome#Finding_A_Time_Server

I don't know if it's legal to deeply link to it on the main article. —Preceding unsigned comment added by 217.235.11.65 (talk • contribs) 18:41, 26 March 2005

Lawsuits
I am surprised no one has mentioned the Law Suits that NTP has on RIM (& in consideration are MS and IBM) —Preceding unsigned comment added by 130.63.241.79 (talk • contribs) 12:41, 24 November 2005
 * That's because this article is about the Network Time Protocol, not about NTP, Inc. RossPatterson 03:33, 3 April 2006 (UTC)

Denial of Service
The University of Wisconsin Madison suffered a denial of service "attack" from flawed NetGear routers using NTP. Seems like something relevant to this article, though the relevance may be more tangental than anything. Perhaps a request for a new article is warranted? I've not done that before...

The history of the NTP DoS at the UW-Madison as told by the source is at http://www.cs.wisc.edu/~plonka/netgear-sntp/ —Preceding unsigned comment added by DavidDouthitt (talk • contribs) 19:03, 8 November 2005
 * I heard this story repeatedly from the horse's mouth. What happened was that Netgear outsourced the production of their routers to China, and the designs done there were subpar. They all targetted the same time server at UW Madison, and the reply packets were by default blocked by the firewall, causing a retransmission every second. Think about that - millions of routers hitting on the same time server once a second (whoa...). Eventually were pouring in excess of 50 megabits of traffic onto the UW network. Mills called netgear to tell them that they needed to stop production immediately and fix the problem; Netgear's response was that they saw no reason why they should. Mills then adviced UW to take netgear to court, but UW decided to work out a deal instead. &rarr;Raul654 00:56, 9 November 2005 (UTC)

This topic is covered in NTP server misuse and abuse. Jaho 02:52, 15 March 2007 (UTC)

Significance of port #123?
Dunno if it's worth mentioning in the main article, but in the UK 123 is the phone number for the Speaking clock. Anyone know if there's any connection between the two? mh. 01:34, 18 December 2005 (UTC)


 * There's the fact that Greenwich is historically the world's time authority, so choosing port 123 would make a kind of sense. --Rfsmit (talk) 19:04, 7 December 2007 (UTC)

Illustration misleading
Under Network_Time_Protocol, the illustration has arrows indication traffic pointing down. This implies that the higher strata servers push time sync data to the lower strata, when in fact it is a pull operation. —Preceding unsigned comment added by Rfsmit (talk • contribs) 18:56, 7 December 2007 (UTC) (edit: Sorry -- adding sig) --Rfsmit (talk) 18:58, 7 December 2007 (UTC)
 * Am I the only person that thinks those arrows are YELLOW? I was looking at the diagram for a few seconds thinking to myself "what green arrows?" until I realised it must mean the yellow ones. --Lytel (talk) 17:33, 2 June 2009 (UTC)
 * I looked at the history of the image, turns out it was change to be colour-blind-friendly. I've updated the description on the page... hope that's OK. --Lytel (talk) 17:34, 2 June 2009 (UTC)

Security Concerns
Several security experts have voiced their concern about the inherent security of NTP. After attacks targeting highly secured servers in 2009 were revealed to have been broken into via an NTP client vulnerability. Numerous figures in the security community have voiced their concerns about the security of the code base and have called for comprehensive review of the protocol and several wide-spread clients. —Preceding unsigned comment added by SteveKostecke (talk • contribs) 23:01, 12 June 2009 (UTC)

This unverifiable clains have been replaced with properly cited text. SteveKostecke (talk) 22:18, 17 June 2009 (UTC)

How is this used with routers
A lot of notes and discussions on the internet about routers and NTP. I can't find any other subject that "NTP" might stand for, so I assume they are referring to this. But how is this protocol significant to router operation?\\


 * Routers only get their timing information from an accurate time source. NTP is the means to transmit that time reference information.--173.69.135.105 (talk) 04:45, 19 November 2011 (UTC)

Source
I don't know where are NTP servers times comes from? Does it comes from computer's clock? Or what?--125.27.51.246 (talk) 19:17, 26 October 2010 (UTC)


 * If I understand it right, the original time source is the U.S. National Institute of Standards (NIST). However, transmission latency can cause some minor errors in accuracy.--173.69.135.105 (talk) 04:43, 19 November 2011 (UTC)
 * NTP is based on Coordinated Universal Time. Several atomic clocks and other stratum 0 devices are synchronized on UTC, but how this is done is outside the scope of NTP.Calimo (talk) 14:58, 21 November 2011 (UTC)

Indefinite article
See Administrators' noticeboard/Incidents and Talk:XMPP. Andrewa (talk) 15:16, 24 May 2013 (UTC)

TCP/IP or not?
NTP is a purely UDP/IP protocol, not TCP/IP.

NTP is one of the oldest TCP/IP protocols still in use

Aren't those two lines contradicting? They are both in the article, so they should either be corrected, or explained —Preceding unsigned comment added by 138.251.200.242 (talk • contribs) 11:12, 4 November 2004
 * That comes from 20040921 edit by 81.242.243.124. I'm wondering if that's a gremlin. He changed "NTP program developed by the OpenBSD project" to "SNTP..." which seems wrong. Someone with better knowledge of protocols should see.--Chealer 14:57, 2004 Dec 3 (UTC)


 * NTP is a TCP/IP protocol (which doesn't mean it has anything to do with TCP). No such thing as UDP/IP exists.
 * Yes. the name "TCP/IP" is confusing. -- Naive cynic 13:00, 20 Jun 2005 (UTC)


 * Terminology was cleaned up at some point to indicate NTP is a UDP/IP protocol. ~KvnG 16:16, 27 May 2013 (UTC)

Network Time Foundation
has removed mention of Network Time Foundation. Regardless of whether Network Time Foundation is independently notable, that they're responsible for ongoing NTP development is a notable fact concerning NTP. I would like to restore this to the lead. ~KvnG 18:33, 10 November 2014 (UTC)
 * I've added this to the History section now that sources have arrived. It looks like the Foundation may lose its funding this month. TwoTwoHello (talk) 19:25, 4 April 2015 (UTC)
 * The refs are not saying that NTF is an effective organization; one suggests the opposite. If the NTF loses its funding, then why should it be mentioned in this article? I do not get a good impression of NTF when I read its IRS Form 990. Glrx (talk) 02:01, 7 April 2015 (UTC)
 * I think I misread "Linux Foundation's $7,000-a-month contribution to NTP" as contribution to NTF. Re-reading the sources, I see the funding is going directly to Harlan Stenn. I agree mentioning the NTF is still not justified (although mentioning Stenn surely is) and unless I can find other sources, I propose to remove "and his Network Time Foundation" from the article. TwoTwoHello (talk) 10:03, 7 April 2015 (UTC)

leap second handling
I believe these sentences are wrong, in general:


 * Because of the requirement that time must appear to be monotonically increasing, a leap second is inserted with the sequence 23:59:59, 23:59:60, 00:00:00. Although the clock is actually halted during the event, any processes that query the system time cause it to increase by a tiny amount, preserving the order of events.

These statements are only partially supported by the cited reference. Showing an inserted leap second as 23:59:60 is what a system honoring UTC is supposed to do, but as far as I know most systems do not. Most systems do not actually have a consistent way of representing the time 23:59:60, so they must use various imperfect workarounds, such as jumping time backwards for a second so that the last second (23:59:59) is repeated, or slowing time down so that the last second takes two seconds. (That is, many systems do not manage to achieve the requirement that time be monotonically increasing across a leap second.)

I believe a correct statement would be


 * Once NTP has determined that a leap second is upcoming, it depends on the operating system in use how the leap second will be represented and displayed. Some systems jump time backwards for a second so that the last second of the day is repeated, others slow time down or stop it so that the last second takes two seconds.  Various additional steps may be taken in an attempt to preserve the necessary appearance that time is monotonically increasing.

and after polishing it a little more, that's the edit I intend to submit. —Steve Summit (talk) 03:40, 17 November 2015 (UTC)


 * I agree that there is no 23:59:60 in NTP. It is also difficult in practice to separate NTP behavior from OS behavior (e.g. Unix time). I think it is still valuable to document NTP time behavior through a leap second. RFC 7164 section 3.4 is an additional reference for this. ~Kvng (talk) 14:29, 20 November 2015 (UTC)


 * Thanks. That's very useful. —Steve Summit (talk) 12:27, 21 November 2015 (UTC)


 * Section now edited along the lines suggested. —Steve Summit (talk) 15:03, 21 November 2015 (UTC)


 * I have reverted your changes. You have deleted the description of NTP time behavior. The reference for the section describes NTP time behavior. If we want to also describe OS time behavior, we need another reference (e.g. RFC 7164). ~Kvng (talk) 14:42, 26 November 2015 (UTC)

Number of strata
The article doesn't explain why "It is hoped that in NTP 5, a protocol still in development, only 8 strata will be permitted." —Preceding unsigned comment added by 82.39.247.48 (talk) 15:56, 11 October 2009 (UTC)


 * Sentence is no longer in the article. ~Kvng (talk) 14:37, 29 September 2016 (UTC)

WikiProject Time assessment rating comment
A basic B as far as WikiProject Time is concerned.

Want to help write or improve articles about Time? Join WikiProject Time or visit the Time Portal for a list of articles that need improving. -- Yamara 19:07, 17 January 2008 (UTC)


 * I have promoted to B-class. ~Kvng (talk) 14:47, 29 September 2016 (UTC)

History
There is no history discussion in this article. Here are a few sources that can be used to write such a section, ,. ~KvnG 19:30, 6 October 2014 (UTC)


 * The history section I added in December 2014 was all sourced (until the final paragraph) to a chapter of, which is available via Google Books. I notice that since then, further sources have been added which is, of course, great. Two sentences have recently been tagged as requiring citation, but I am not sure why. Is it because the source is not available to some readers? or perhaps because I should have repeated the source at the end of every paragraph? (every sentence?) I notice that the second source listed above supports the content in question. Would that source be preferred to the book? TwoTwoHello (talk) 10:53, 11 November 2017 (UTC)

NTPsec new implementation of NTP
NTPsec is a secure version of NTP and a project fork of the reference implementation. They should probably be included in the page. You can find information on them here. TMLutas (talk) 21:32, 31 January 2016 (UTC)


 * Here is some coverage. Already mentioned at ntpd. Seems like it is worthy of inclusion here too. ~Kvng (talk) 14:42, 29 September 2016 (UTC)

Tech lead of NTPsec here. The summary of the PenTest report is, I feel, somewhat misleading. There were two new minor defects identified, but it would be fair to add that no CVE was issued for either as they were not found to be exploitable. I won't do the edit myself, but I request that one of the regular page maintainers consider it. Eric S. Raymond (talk) 20:18, 26 August 2019 (UTC)


 * Thank you for looking over the NTPsec and offering suggestions for improvement. On page 6/18, near the top, of the PenTest report, CVE-2014-9295 was identified as being introduced to NTPsec through a regression and they state "the code now introduces an authenticated DoS." So while a CVE defect may not have been announced, they found and reported on at least one CVE in the code. I am not an expert and am possibly missing something here. But the assertion that there were no CVEs at the time of the report seems inaccurate. Comments? --  22:12, 26 August 2019 (UTC)

I'll check with our security officer about that. He keeps our NEWS file updated with CVEs we've dodged and (supposedly) any CVEs issued against our code. I just looked and CVE-2014-9295 isn't there. I will be displeased if it turns out to have been erroneously omitted. Thanks for the response. Eric S. Raymond (talk) 05:31, 27 August 2019 (UTC)

Put some image in the lead?
Right now, the first image of the article is File:DL Mills-2.jpg. That causes this photograph to be displayed as the image of the navigation popup, and I presume the page preview as well. I have nothing against the photograph but it is a bit weird to have a person as a preview for such a subject.

How about bringing File:Network_Time_Protocol_servers_and_clients.svg (already present in the article) into the lead to solve this? Tigraan Click here for my talk page ("private" contact) 14:56, 17 December 2021 (UTC)


 * David Mills is kind of singlehandedly responsible for bringing NTP into being. OTOH your suggested image does a good job of graphically summarizing how the protocol works. Probably the best thing to do is add Infobox protocol to the top of the article with your diagram featured there. ~Kvng (talk) 15:34, 20 December 2021 (UTC)

32-bit fractional second part
So you multiply the fractional part by 2-32 then add it to the integer part to get the time in seconds? Ojw 14:43, 30 August 2005 (UTC)


 * Yes ~Kvng (talk) 16:02, 1 February 2022 (UTC)

Chrony NTP daemon
Please mention Chrony as well as Ntp and Openntpd. The home page for the project is http://http://chrony.sunsite.dk/

No need to. This article is about the NTP protocol, ntpd and OpenNTPD are the major deamons implementing it. There are dozens NTP clients out there and chrony is just one of them. Jaho 20:01, 15 March 2007 (UTC)


 * chrony, OpenNTPD and ntpd each now have their own articles. All are also described in Network_Time_Protocol. ~Kvng (talk) 16:05, 1 February 2022 (UTC)

Clarification needed?
In the "NTP Timestamps" section, there is the following quote:

"The 64 bit value for the fraction is enough to resolve the amount of time it takes a photon to pass an electron at the speed of light."

Unless I'm misunderstanding something, isn't there something missing here? What is the initial distance between these two particles (or whatever they are)? If they're billions and billions of light years apart, for example, then I could probably "resolve the amount of time" for one to pass the other using my kitchen clock. Is this an incomplete quote or am I just confused? 65.183.135.166 (talk) 21:20, 28 November 2007 (UTC)

I think he's trying to refer to the amount of time that a photon would take to travel the diameter of an electron. I'm not sure the quote is clear and even if it is I'm not sure anyone really knows how wide an electron is? There is a classic radius O(10^-15)m, perhaps that is what is being referred to? 92.251.70.206 (talk) 06:13, 5 June 2008 (UTC)

(The point seems to be that 64+64 bits is sufficient for any possible timekeeping needs in our universe, in a stronger sense than the prediction that [|"nobody will ever need more than X kilobytes of memory"]. 71.139.177.112 (talk) 01:17, 16 February 2009 (UTC)


 * This is a quote from David Mills. He doesn't seem to have the physics quite right. There's now a footnote there that relates 2-64 s to the Bohr radius. ~Kvng (talk) 16:11, 1 February 2022 (UTC)

Another clarification needed
The section also says

"Implementations should disambiguate NTP time using a knowledge of the approximate time from other sources. Since this only requires time accurate to a few decades, this is unlikely to ever be a problem in general use."

I think I see what this means, but I'm not sure. Maybe an example and an explanation of the assumptions would clarify it, as in:

"For example, an operating system version that was released in the early 21st century can safely assume that the system clock should never be set to a time in the 20th century, and can probably assume that it will not still be in use in the 22nd century. In this case, a time built into the operating system would be the other source."

But it could equally well be talking about the DAYTIME protocol, file timestamps, or the computer's hardware clock. Someone who is confident of what is intended should disambiguate it.

--71.139.177.112 (talk) 01:17, 16 February 2009 (UTC)


 * No longer in the article. Instead it talks about the 128-bit date format introduced in NTPv4. NTP experts seem unconcerned about the rollover issue confident it can be disambiguated. I don't know if this confidence is warranted or hopeful thinking. ~Kvng (talk) 16:17, 1 February 2022 (UTC)

Requesting community opinion on archiving older parts of this talk page
I would like to archive (and/or hava a bot archive) older parts of this talk page. A lot of the sections have not had any discussion for some time now. Many of these are support requests and I think that removing them might help to limit people to resist adding further ntpd support requests. I am not sure how exactly to establish community consent, please indicate if you have any objections.

DouglasCalvert (talk) 06:40, 24 May 2013 (UTC)
 * Why not just remove the inappropriate comments per wp:notaforum? In fact, I am going to have a go at some of them now. The talk page doesn't seem busy or large enough to justify archiving yet. TwoTwoHello (talk) 09:18, 24 May 2013 (UTC)
 * I have set up automatic archiving. ~Kvng (talk) 16:19, 1 February 2022 (UTC)

the root daemon?
my bullshit detector just got triggered: in the "security concerns" section we read "that can lead to unlimited access to systems that are running some versions of NTP in the root daemon. Systems that do not use the root daemon, such as BSD, are not subject to this flaw"

There is no such thing as "the root daemon" in unix system. While we can muse philosophically about things like init and systemd, none of this is relevant to the NTP topic, and the gicven source also does not make any such claims in the first place. Who writes this bullshit, and what does he want to tell us? If noone objects ill rephrase in more standard Unix terms, according to the (somewhat dated) source Wefa (talk) 18:41, 17 May 2017 (UTC)
 * Please go ahead and rephrase it. --LiberatorG (talk) 06:58, 18 May 2017 (UTC)
 * It doesn't look like this has been addressed. I have added a dubious tag pointing to this discussion. ~Kvng (talk) 16:24, 1 February 2022 (UTC)

Reference clocks
Hey, where did you get this material? Are we sure this is the best place for it. I was thinking it might be better in Master clock. ~Kvng (talk) 14:39, 4 April 2018 (UTC)


 * Material is no longer in the article. ~Kvng (talk) 16:27, 1 February 2022 (UTC)

First paragraph of Security section needs updating
In the first paragraph of the security section it says:


 * As of January 2011, there are no security revisions in the NTP specification and no reports at CERT.

Given that it is 2020, we should either update this sentence with newer info or remove it. I don't have the time today to do the research but am flagging it here in case someone else has a chance before I get the time to do so. - Dyork (talk) 20:54, 7 May 2020 (UTC)


 * The most recent date in Network_Time_Protocol is now 2017 and there are no "as of" statements. Seems durable. ~Kvng (talk) 16:30, 1 February 2022 (UTC)