[LEAPSECS] LEAPSECS Digest, Vol 45, Issue 15

Jonathan Natale jnatale at juniper.net
Sat Sep 4 16:03:34 EDT 2010


AFAIK, most, deployed NTP implementations on routers, if not in general, simply ignore leap seconds and fall slightly out of sync for a bit. AFAIK, NTP on routers, if not in general, use NTP mainly to sync logs across various boxes, and usually being 1 second off is not a big deal. Ditto for the NTP eras--it doesn't fly and nobody cares (we'll fix it when it breaks, or we'll be dead by then anyway). The issue is the OSs--no "xx:xx:60" support.

What is wrong w/ the "http://www.cl.cam.ac.uk/~mgk25/time/utc-sls/draft-kuhn-leapsecond-00.txt" sol'n? Seems simple and effective enough for most (definitely not for all) apps.

-$0.02

-----Original Message-----
From: leapsecs-bounces at leapsecond.com [mailto:leapsecs-bounces at leapsecond.com] On Behalf Of leapsecs-request at leapsecond.com
Sent: Saturday, September 04, 2010 10:20 AM
To: leapsecs at leapsecond.com
Subject: LEAPSECS Digest, Vol 45, Issue 15

Send LEAPSECS mailing list submissions to
leapsecs at leapsecond.com

To subscribe or unsubscribe via the World Wide Web, visit
http://six.pairlist.net/mailman/listinfo/leapsecs
or, via email, send a message with subject or body 'help' to
leapsecs-request at leapsecond.com

You can reach the person managing the list at
leapsecs-owner at leapsecond.com

When replying, please edit your Subject line so it is more specific than "Re: Contents of LEAPSECS digest..."


Today's Topics:

1. Re: Coming of age in the solar system (Paul Sheer)
2. Re: Coming of age in the solar system (M. Warner Losh)
3. Re: Coming of age in the solar system (M. Warner Losh)
4. Re: Coming of age in the solar system (Ian Batten)
5. Re: Coming of age in the solar system (Poul-Henning Kamp)
6. Re: Coming of age in the solar system (Rob Seaman)
7. Re: Coming of age in the solar system (M. Warner Losh)
8. Re: Coming of age in the solar system (Paul Sheer)


----------------------------------------------------------------------

Message: 1
Date: Sat, 04 Sep 2010 13:44:22 +0200
From: Paul Sheer <p at 2038bug.com>
Subject: Re: [LEAPSECS] Coming of age in the solar system
To: Leap Second Discussion List <leapsecs at leapsecond.com>
Message-ID: <1283600662.9314.20.camel at localhost>
Content-Type: text/plain



>

> Of course, it would be cheaper for the software folks to never have to

> worry about it again. That would also make them predictable.

>


The software industry doesn't know or care about leap seconds.

The whole time issue is universally dealt with using a single line in all installation manuals:

"We recommend customers install NTP."

The few specialized applications that need to worry about leap seconds have long since worked around the problem.

-paul







------------------------------

Message: 2
Date: Sat, 04 Sep 2010 06:22:20 -0600 (MDT)
From: "M. Warner Losh" <imp at bsdimp.com>
Subject: Re: [LEAPSECS] Coming of age in the solar system
To: leapsecs at leapsecond.com, seaman at noao.edu
Message-ID: <20100904.062220.475505532728779189.imp at bsdimp.com>
Content-Type: Text/Plain; charset=us-ascii

In message: <FB1295B4-A884-4984-8A7A-CA46F9586952 at noao.edu>
Rob Seaman <seaman at noao.edu> writes:

: On Sep 3, 2010, at 11:52 PM, M. Warner Losh wrote:

:

: > And I'd be pointing to those same discussions where actual practitioners discuss the difficulty in the phrase "just an issue of representation."

:

: I believe my boss expects me to actually practice in exchange for my

: salary :-)


The problem is that your expectation of what my boss wants differs from what my boss wants. In general, development management doesn't care about esoteric edge cases like leap seconds, and would rather have developers spend the time on more important things.


: > There's both too much automation in leap seconds, and not enough. Nobody cares enough to solve the whole problem

:

: And right now I'll do the one thing that seems to drive people

: craziest - suggest that system engineering best practices be

: followed to first understand the whole problem - before pursuing a

: solution actually responsive to the requirements. (What a concept!)


Again, you are missing the point: Your engineering solution isn't what is driving software development. The engineering trade offs you'd make are different than the ones others would make. It is rare that you account for absolutely everything in the design, and leap seconds are often sacrificed to make the job fit within the cost constraints in place.


: > I've said many times: leap seconds happen to unpredictably to be on the radar of most people.

: >

: > If they were on a regular schedule, where on the average we tried for a |DUT1| < 10s, (or 5s or whatever) we'd have a better chance of getting them right.

:

: And folks on this side of the aisle have expressed willingness to

: explore these possibilities. Such options were mentioned in the

: ancient Dead Sea Scrolls - er, GPS World article. Why then has the

: only option entertained by "the deciders" been to kill leap seconds

: dead, dead, dead?


I don't know. But I have no control over what other people entertain as thoughts.


: > Sadly, there's no serious discussion of this middle ground outside of this list.

:

: So shouldn't we attempt to expand the discussion? Getting the original archives back online would be a good start...and then look into organizing a meeting to engage with key stakeholders - many of who will have never even heard of the ITU proposal (or the ITU for that matter).

:

: > Of course, it would be cheaper for the software folks to never have to worry about it again.

:

: There's that assumption again - leap seconds are deadly, but

: absolutely nothing bad can come from redefining every single clock

: on the planet.


That would be an argument against leap seconds too. When they were introduced, they obsoleted all mechanical clocks because none of them can handle leap seconds. Even today, the number of clocks that can handle them is vanishingly small.

Warner



: On Sep 4, 2010, at 12:11 AM, Poul-Henning Kamp wrote:

:

: > In message <6C773408-836C-4D2A-9D37-0A8B7111E675 at noao.edu>, Rob Seaman writes:

: >> The flip side of timekeeping on a spinning pebble skipping through the cosmos is that nothing prevents *also* benefitting from an atomic timescale devoid of the dreaded irregular radix.

: >>

: >> Civil time = mean solar time

: >> Techie time = atomic clock ensemble

: >

: > Absolutely, and that is what people are advocating in various forms.

: >

: > The issue that you in particular, and some astronomers along with you, refuse to acknowledge, is that "techie time" extends all the way to peoples telephones, internet connections, flight safety, medicine production and so on.

:

: I hereby acknowledge that techie time extends all that long, long, way. And that these are all systems connected to networks. And that those networks are needed to run NTP in the first place. And that the same networks that convey UTC can *also* convey TAI (or some duly blessed transportable realization of atomic time) - or say - hey! - if only there were a handheld device that could magically receive a time signal from outer space, a time signal devoid of those deadly Kryptonite leap seconds.

:

: > Civil time these days would be more correctly called "ornamental time", because it starts at the VCR, fridge, oven, grandfather clock and other off-line time displays.

:

: And there are archived discussions pointing out the irrelevancy of such systems to this discussion.

:

: Certainly civilians have need ("use cases") for both civil time as well as techie time. UTC manages the neat trick of serving both. Don't like that? Split the two. Oh wait! GPS already did that! But oh :-( the Timelords have an even greater distaste for GPS than for UTC...and they also appear to be poised to kill off TAI. This raises the question of why they got in the business of timelordliness in the first place since they don't appear to like any of the many, many, timescales.

:

: Three birds in the hand are certainly worth a whole lot more than one in a very thorny shrubbery.

:

: Where are the Knights who say Ni! when you need them?

:

: Sir Rob is running away now...

: _______________________________________________

: LEAPSECS mailing list

: LEAPSECS at leapsecond.com

: http://six.pairlist.net/mailman/listinfo/leapsecs

:

:



------------------------------

Message: 3
Date: Sat, 04 Sep 2010 06:25:22 -0600 (MDT)
From: "M. Warner Losh" <imp at bsdimp.com>
Subject: Re: [LEAPSECS] Coming of age in the solar system
To: leapsecs at leapsecond.com, p at 2038bug.com
Message-ID: <20100904.062522.821837081338301975.imp at bsdimp.com>
Content-Type: Text/Plain; charset=us-ascii

In message: <1283600662.9314.20.camel at localhost>
Paul Sheer <p at 2038bug.com> writes:

:

: >

: > Of course, it would be cheaper for the software folks to never have to

: > worry about it again. That would also make them predictable.

: >

:

: The software industry doesn't know or care about leap seconds.

:

: The whole time issue is universally dealt with using a single

: line in all installation manuals:

:

: "We recommend customers install NTP."

:

: The few specialized applications that need to worry about leap

: seconds have long since worked around the problem.


That's exactly the attitude that keeps leap seconds from working properly. Oh, just do X and there will be no problems. Of course, by saying that, you still have the problem. NTP doesn't solve the leap second problem: you still have weird stuff that happens at the leap second. NTP usually lets you recover from it, since the last few leap seconds have shown that even today's modern ntp network can't get the leap seconds right.

This gets back to my earlier statement: When the definition of right is "didn't crash the OS" and not "ticked time correctly" you get all kinds of wacky behavior.

Warner


------------------------------

Message: 4
Date: Sat, 4 Sep 2010 14:12:25 +0100
From: Ian Batten <igb at batten.eu.org>
Subject: Re: [LEAPSECS] Coming of age in the solar system
To: Leap Second Discussion List <leapsecs at leapsecond.com>
Message-ID: <060DD241-6C0D-4A8E-8E12-42C5FE198361 at batten.eu.org>
Content-Type: text/plain; charset=US-ASCII; format=flowed; delsp=yes


> Of course, by

> saying that, you still have the problem. NTP doesn't solve the leap

> second problem: you still have weird stuff that happens at the leap

> second.


But no-one, for an engineering value of no-one, cares. Once every
few years, hourly statistics will be off by 0.03%. Where's the pressure coming from to claim this is an issue?

ian


------------------------------

Message: 5
Date: Sat, 04 Sep 2010 13:35:17 +0000
From: "Poul-Henning Kamp" <phk at phk.freebsd.dk>
Subject: Re: [LEAPSECS] Coming of age in the solar system
To: Leap Second Discussion List <leapsecs at leapsecond.com>
Message-ID: <93160.1283607317 at critter.freebsd.dk>

In message <1283600662.9314.20.camel at localhost>, Paul Sheer writes:


>The whole time issue is universally dealt with using a single line in

>all installation manuals:

>

> "We recommend customers install NTP."

>

>The few specialized applications that need to worry about leap seconds

>have long since worked around the problem.


You should really read the archives, before you say something as stupid as that.

--
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG | TCP/IP since RFC 956
FreeBSD committer | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.


------------------------------

Message: 6
Date: Sat, 4 Sep 2010 06:45:10 -0700
From: Rob Seaman <seaman at noao.edu>
Subject: Re: [LEAPSECS] Coming of age in the solar system
To: Leap Second Discussion List <leapsecs at leapsecond.com>
Message-ID: <BBCA1C8E-8305-4B17-8A38-0B50A5BF5EF7 at noao.edu>
Content-Type: text/plain; charset=us-ascii

On Sep 4, 2010, at 6:35 AM, Poul-Henning Kamp wrote:


> You should really read the archives,


Are the archives from the original list available somewhere? I would appreciate a pointer!

Rob



------------------------------

Message: 7
Date: Sat, 04 Sep 2010 07:59:50 -0600 (MDT)
From: "M. Warner Losh" <imp at bsdimp.com>
Subject: Re: [LEAPSECS] Coming of age in the solar system
To: leapsecs at leapsecond.com, igb at batten.eu.org
Message-ID: <20100904.075950.18103541983818278.imp at bsdimp.com>
Content-Type: Text/Plain; charset=us-ascii

In message: <060DD241-6C0D-4A8E-8E12-42C5FE198361 at batten.eu.org>
Ian Batten <igb at batten.eu.org> writes:

: > Of course, by

: > saying that, you still have the problem. NTP doesn't solve the leap

: > second problem: you still have weird stuff that happens at the leap

: > second.

:

: But no-one, for an engineering value of no-one, cares. Once every few

: years, hourly statistics will be off by 0.03%. Where's the pressure

: coming from to claim this is an issue?


That's rather my point: NTP doesn't get things right, and there's an attitude that it doesn't matter since things are mostly papered over.
Even without NTP things are mostly right, but nobody cares enough to get it right. So almost nothing gets leap seconds right, and we're left with a lot of warts in the system because of it.

It is because of the 'good enough' attitude that doing any better is almost impossible for those applications that do care.

Warner


------------------------------

Message: 8
Date: Sat, 04 Sep 2010 16:18:06 +0200
From: Paul Sheer <p at 2038bug.com>
Subject: Re: [LEAPSECS] Coming of age in the solar system
To: Leap Second Discussion List <leapsecs at leapsecond.com>
Message-ID: <1283609887.9314.51.camel at localhost>
Content-Type: text/plain

On Sat, 2010-09-04 at 07:59 -0600, M. Warner Losh wrote:

> In message: <060DD241-6C0D-4A8E-8E12-42C5FE198361 at batten.eu.org>

> Ian Batten <igb at batten.eu.org> writes:

> : > Of course, by

> : > saying that, you still have the problem. NTP doesn't solve the

> leap

> : > second problem: you still have weird stuff that happens at the

> leap

> : > second.

> :

> : But no-one, for an engineering value of no-one, cares. Once every

> few

> : years, hourly statistics will be off by 0.03%. Where's the pressure

> : coming from to claim this is an issue?

>

> That's rather my point: NTP doesn't get things right, and there's an

> attitude that it doesn't matter since things are mostly papered over.

> Even without NTP things are mostly right, but nobody cares enough to

> get it right. So almost nothing gets leap seconds right, and we're

> left with a lot of warts in the system because of it.

>

> It is because of the 'good enough' attitude that doing any better is

> almost impossible for those applications that do care.

>



Well...

I also thought it was a huge problem.

So I took to a casual strolling down the hall to our support team, and did thusly spake: "Hay righteous $upport dudes: do you know about this here leap second thing?"

And they did sayeth: "yep"

And I did spake: "Isn't it, like, a huuuuge problem for our customers?"

And they said: "you mean the 150 telecoms operators that service over half a billion subscribers betwixt them?"

Me: "yeah those!"

And they said: "no"

And I did retort: "has it ever been a problem, or have you ever had a customer effected by leap seconds?"

And they said: "no"

Me: "and what other vendors have you all worked at in the course of your esteemed careers in the telecoms industry?"

Support team replied with the names of all the large vendors of telecoms equipment.

Me: "and has it ever been a problem for anyone ever?"

And they said: "no"

And that was the end of the discussion.

-paul









------------------------------

_______________________________________________
LEAPSECS mailing list
LEAPSECS at leapsecond.com
http://six.pairlist.net/mailman/listinfo/leapsecs


End of LEAPSECS Digest, Vol 45, Issue 15
****************************************


More information about the LEAPSECS mailing list