Poor timekeeping on a RasPi-5/4GB after Bookworm OS update....

r /Commercial IT Users/Software engineers/

The other class of computer users who'd likely use a screen the same way as a software engineer or developer is an author (of anything from scripts to novels).

I was usually able to get access to a keypunch, although sometimes it meant waiting until the keypunch staff was off for lunch. In a pinch I did use one of those hand punches; I affixed a tag to it saying "Programmers have priority on this punch!"

I never did use paper tape for programming, but many of our customers sent us data on paper tape so I became the local guru, even doing several re-writes of the IOCS to get the desired results.

Being a denizen of the mainframe world, I didn't get to play with teletypes much - at least in a work environment.

Ah yes, I remember mainframe terminals and their programming nightmares. I did manage to port Dungeon to that environment, which took some doing.

Ditto.

When I start getting weird about non-computing things, my wife says, "Why does everything have to be such a huge project?" But with IDEs, everything _does_ have to be a project. Makefiles are so much simpler.

A couple of times I've fought bugs that were sufficiently obscure that my usual practice of sprinkling printf()s (or, more accurately, writes to a log file) through the code was progressing painfully slowly. I tried loading my code into an IDE, but it turned out to be so much work setting up those damned projects that I found it easier to give up on IDEs and tough it out the old-fashioned way.

Having said that, there are times when I find a GUI more convenient - but more often than not a good old CLI does the job much more quickly and easily. I reserve the right to use whatever method works best at the time, and resent any attempt to force me into a mold, however pretty the graphics may be.

We used 12 key punches in the first programming shop I worked in, and yes, we hand keyes patches as well as complete programs if our keypunch pool was busy (this was a small computer bureau).

The first computer I used was an Elliott 503 (scientific beast, programmed in ALGOL 60). We used a Flexowriter to write and edit programs on 8 column paper tape.

operators console - they didn't use operator consoles with screens: those were only fitted to 2900 series machines.

So what were you on - IBM, Burroughs or what?

Indeed, though the ICL trick of locally editing the current page was relatively user-friendly, but I never did like the way the OS/400 editor (on IBM AS/400 systems) worked.

Quite - thats why I like to drive C and Java program builds and don'y like the IDE's

Yep nothing much beats a command line and a well written makefile, even it its an 'ant' build script for Java: thre capabilities are quite similar, though an ant build file uses less arcane syntax than make.

I started in a small service bureau as well. Fortunately I was usually able to get my hands on a keypunch; the keypunch department used Univac 1710s but we had an IBM 29 for small jobs. I could punch my own cards as fast as the keypunch girls (they weren't used to alphanumeric stuff). I avoided coding forms like the plague; I had all kinds of shorthand that nobody would understand and could improvise on the fly.

Univac. My first machine was the 9300, which was their answer to the IBM 360/20. No console typewriter, just front-panel lights on which you'd display a pattern which the operator would read in hexadecimal and look up in a big book. Originally it was a pure card system with 16K of memory, but later we added a couple of 8414s (clones of IBM 2314 spindles) and increased the memory to a whopping 32K. We didn't know what to do with all that space (although we soon figured that out).

I did some work on 9400s, then was involved in the 90/30 and its follow-ons, running OS/3, for about 10 years.

Univac's serial port protocol was a beast. It was supporting multi-dropped block-mode synchronous terminals, and the handler program (ICAM, which stood for Integrated Communications Access Method) was a nightmare to generate. Even later, when I worked for Univac and was supposedly one of the gurus, our standard way of building a new installation was to copy an existing one and tweak it as required. As far as we knew, there was only one person in all of Canada who really understood it, and we flew him out from Montreal a couple of times to help us out. After one such session at a customer site, we were sitting around shooting the bull, and he was scrambling and unscrambling a Rubik's Cube while keeping up with the conversation. That's the kind of mind it took to understand ICAM.

GPS *is* more precise, no question. However few NEED that level of precision. A few seconds accuracy is more than enough. Time servers do that and a bit better - but ONLY a bit better.

First off ... 5ghz is MUCH worse at penetrating common structures. Everything I'd been doing still relied on 2.4 ... easy 25%/50% better signal.

And, as said, do you REALLY need micosecond+ accuracy ?

Beyond microsecond you're dealing with speed-of-electricity/ speed-of-light issues and Einie sez you're screwed. Only dedicated scientific devices do better - and their timing is RELATIVE, only for THAT device. Consider the LIGO grav "telescope" as an example. It's not using NTP to do its work.

Ah ... but writing TEXT ... and "Word-Processing" ... are entirely different things.

Yep, you can do TEXT with nano or maybe something a tad better. But to craft an appealing DOCUMENT is far and above that level. Docs have to LOOK GOOD. Nano ain't gonna do that. LibreOffice instead these days .......

Actually, I *remove* Vi and its ilk from every distro I install. HATEFUL apps. Never EVER wanna see them - as bad as "edlin" from the early DOS days.

"Mail" is not "word processing" either. Generally it's pure TEXT. If it's anything important you do it with a REAL word-processor and then mail a PDF attachment.

Good old ASCII has it's place - but ONLY a place.

Heh ! I saw that recently in a US-based bank ! The moment you depart from the "usual" it's back to mostly terminal-based apps. They're 25 years behind the curve and I don't think it's gonna get any better, given modern money constraints.

The "golden age" was the mid 60s/early-70s. The USA was fat with money back then. Banks/corps/govt could afford to have serious apps (usually COBOL) custom-writ. GOOD CODE too, by the white-shirt/narrow-tie guys (few girls then). They STILL use these apps because they can't AFFORD anything newer now.

IT people who are mostly interested in high/steady income should learn to rock FORTRAN/COBOL because SO much of the vital infrastructure STILL runs on that.

But, if you want "word processing" you do NOT use a terminal- based app. Docs need to LOOK GOOD and even the poorly-trained need to be able to make docs that LOOK GOOD, quick. That's LibreOffice and a few friends.

Hmm ... of note ... some big US banks are cutting WAY back on actual HUMAN assistance. It's mostly the ones who have over-extended themselves on loans, mostly for inflated properties, and expect massive defaults like not SO long ago (people/corps learn NOTHING). "AI", well, it just does not GET "nuances" and is NO SUBSTITUTE.

In the USA there's something called the "ADA", "Americans with Disabilities Act". In truth, anyone over 65 DOES have some disabilities - understanding, even SEEING, those damned phone apps all the banks want everyone to use for everything. There are also LOTS of lawyers who specialize in ADA issues. MAYbe it's time to sue the s*it out of those asshole banks ?

I'm over 65.

My "phone" is a FLIP ... tiny tiny screen. You CANNOT fill out complex forms with it. I am *not* gonna buy anything bigger/more-complex. "Phones" are for PHONE CALLS in my view of things. I don't even like "texts". And the SECURITY for phone apps - APPALLING !

When I see word processor I think what a food processor does.

That's what groff is for - I far prefer markup to wysywig.

I've been using vi for a long time - it's a great text editor.

Who mentioned email ?

If it's important I do it with vi and groff to generate a PDF or print a letter.

Yep UTF-8 is far more useful.

Just why I tried comparing 2.4 and 5 GHz, but it made no difference in this particular case. I didn't expect it to, but it was worth a test.

We've found in the past that "second" level timing is not good enough if, for example, you are comparing log files between different locations, and there are dozens of entries per second. Another use if where you are providing sounds from separate devices, for example voices in a choir, The ear/brain can detect differences of under 100 ms.

In this case, the question is still unanswered - why did an OS upgrade ruin what was excellent timekeeping?

There is the tool ntpmon included with ntpsec. Try it.

A question: Can you show the MRTG setup to produce the curves linkled to on the first post?

Indeed. Viz my =issues with trying to design some extremely simply PCBS for my PI projects. Traditionally I did that with colored tapes pasted on transparent backings. These days you simply need Gerber files. Now te way you are expected to make THOSE is to create a huge 'project' add all youur components to a nassive database of 'components' in eagle CAD, create a circuit diagram, and then manipulate a massively complex program to create multilayer boards. with solder masks vias, silk screen and copper.

Total f****ng OVERKILL. I managed to find an online site that would convert Corel Draw files to Gerber files, and a linux program that can turn a Gerber drill fill into the format CNC machines use so I designed my PCBs in Corel draw. In far less time than it would have taken to learn Eagle CAD.

My point being that I could have been on a training course to learn e,g,. Eagle CAD and become proficient in the tool, but it would not have made the product any better in the long run.

In the end there needs to be an element of intelligent design rather than random monkey evolution for the best results. Darwin does not ensure the survival of the fittest, merely the elimination of the totally crap.

And the problem is that if the Darwinian approach to design predominates, what you get in the end is *almost* complete crap. Good enough to sell, but not to actually work.

Viz Barclays Bank are still sending me messages and statements from a share trading account I transferred all the money out of, two years ago, and asked them to close.

I assume that the software to actually close an account was simply never written.

There is nothing to be done about all of this. The vast majority of humanity is barely capable of reproduction. They don't think. They don't want to think, and if they do think they want to be told by the current political and moral authority *what* to think.

Only in exceptional cases will they learn *how* to think. And my teaching people to read, and developing mass media, we have ensured that everyone now feels they are an expert in everything, and fully capable of understanding everything and their opinion is as good as anyone else's when in reality they haven't a f****ng clue.

I give you Angela Rayner....

- Bertrand Russell

Thanks, Tauno!

I'll take a look at ntpmon which is a tool I've not used before.

The MRTG setup is described here:

I've also seen that at the local Canadian Tire. But why is that not better? Better than what? If the information you need can be nicely presented on a 24x80 green screen, why not do so? It's FAST - which is only to be expected if you're sending 1920 characters instead of several megabytes of HTML and JavaScript that needs to be interpreted. And you don't have to put up with the current fad of light grey characters on a white background.

Don't forget governments. They're doing it too. It's as if having a smartphone is becoming mandatory. It's just about time for a class action suit.

I'm with you. I'll give up my flip phone when they pry it from my cold dead fingers. I want _buttons_, dammit! And I'm amazed at how people can be disturbed by the concept of "ankle bracelets" when a GPS-equipped smartphone is the same thing.

Depends ... some structures have more brick and metal in them and 2.4 is superior. Naught but studs and dry- wall ... take yer pick. In either case, 5ghz only seems to offer a real advantage in like a big open office space - with people rather close to the WAP.

"Second-level" is NOT good enough. However millisecond-level is under almost every use. Microsecond and finer ... nothing to be gained unless you're running 'scientific' devices or maybe extreme cluster-computing.

Ex-M$ idiots now working for Linux distro shops ??? Apparently the highest concentration at Canonical ... dropped THEIR crap a few years ago.

Recently re-did a number of boxes/devices to BookWorm. Now I'm gonna "upgrade" them - back to BullsEye. Some of the Arch distros look good now ... and even some of the BSDs are looking decent these days. We'll see what else is out there ... something "conservative" ....

NTP *can* deliver pretty damned good timing - assuming the daemons are given enough priority and updates happen pretty frequently. The two 'fixes' are GPS - as I think you or someone noted - and/or/with one of those high- precision timing chips. Look up "ChronoDot" ... they are used a lot for microcontroller apps because they are protected against temperature drift and such.

Hmm .. for fun ... find out where the NTP daemon is controlled. Maybe they DID cut its priority way back ? Firewalls can also mess with NTP because there does exist evil-ware that can do sneaky things with it. I'd suggest filtering that only allows a few govt NTP servers in. On a Pi I think you can actually SPECIFY the desired NTP server ... in 'raspi-config' or the GUI version.

That's because one is an alias for the other.

But that is only telling you that the daemon is running but not what it is doing. Use "ntpq -pn" to query what peers it is using and if the daemon can reach them in the first place.

Stefan,

Thanks for your comments.

Here's the ntpq -pn output. RasPi-28 is the problem device, RasPi-24 is an RPi-400 similarly configured:

C:\Users\win-8>ntpq -pn raspi-28 remote refid st t when poll reach delay offset jitter ==============================================================================

-109.74.206.120 129.69.1.153 2 u 34 64 377 20.828 +1.405 0.668

C:\Users\win-8>ntpq -pn raspi-24 remote refid st t when poll reach delay offset jitter ==============================================================================

-162.159.200.123 10.85.8.67 3 u 48 64 377 16.660 +0.123 0.982

-95.97.2.35 192.168.10.25 2 u 23 64 377 37.744 -1.863 1.254

-162.159.200.1 10.85.8.67 3 u 15 64 377 16.701 +0.192 1.457

-178.79.188.22 125.199.53.76 2 u 9 64 377 20.280 +3.615 0.677

-176.58.109.199 131.188.3.223 2 u 27 64 377 21.648 +6.233 2.305

I wish I could use regular NTP and not NTPsec!

I'd disagree. "Documents" and "text" really aren't the same things, for the same purposes.

Yuk !

WORK fine ... but for PUBLICATION ???

It's crap - just a far less friendly version of nano.

So I always REMOVE it so it can't pop up by accident.

I think it was in the thread somewhere ... or at least implied by "text". Of course modern e-mail is mostly HTML these days ....

But you said "word-processors" were EVIL :-)

And, frankly, being able to make nice text and insert photos into e-mail these days is very convenient.

I make PDFs with LibreOffice.

Again YUK. ASCII should have remained ASCII and UTF and friends implemented separately.

DO miss those funky 128+ graphics chars that were in the old IBM roms. Made it so easy to make text-based forms and such. They still exist in UTF, but at very obscure values and with the added 2-byte complications.

It doesn't look too bad. Both devices use your local stratum-1 servers as candidates (+ in first column) and have selected 192.168.0.20 as reference (* is first column). All candidates were reachable for the last 3.7 minutes (you poll about every 32 seconds and 377 indicates the last 10 polls were successful). As you noticed it is really only the jitter that looks suspicious.

I would check this output from time to time and look for anything strange. Could the network in between be an issue?

On raspi-28 you only have the UK pool while raspi-24 also lists the NL pool as reference. That is probably the reason for the different numbers of servers.

You could also try out the chrony NTP client which is mostly used on Enterprise Linux distros. The package is also available on Debian.

Nitpick - eight polls. The reach figure is an eight bit shift register displayed in octal, every poll shifts a 1 (success) or 0 (fail) onto the low end.