The computing stuff rant thread

[chrisbainbridge]

I have a new iPhone on iOS 17 and have noticed minor problems with other Bluetooth connections as well

[Afasoas]

AAISP customer, so IPv6 is normal and ordinary and has been working fine here for years.  As such I get a bit resentful when people's default solution to network probems is "disable IPv6".

I do have a special IPv4-only VLAN/SSID for whichever buggy old Android we had that couldn't cope with RAs (the more recent ones are fine), which also has 802.11b enabled for testing.



I upgraded our desktops from bullseye to bookworm last week.  No drama at all for barakta's.  Mine got some mild font-rendering shenanigans, and I had to re-write my script that toggles between monitor speakers and headphones, because the sound system has been completely replaced.  Again.

Dist upgraded the old daily driver laptop to Bookworm last night. The only issue thus far is a change to the location of the NTP config file and some configuration in it which no longer works. The latter had the knock-on effect of breaking Network Manager when it was trying to do DHCP things to the running NTP configuration.

I've got one Android device left which doesn't play nice with IPv6. When that dies, I'll experiment with turning router advertisements back on. I need to set-up some IPv6 VPN gateways and adjust my home-brew Pi-hole equivalent (a set of scripts that periodically update Bind response polizy zones) to also blackhole nefarious/spammy IPv6 lookups.

[Afasoas]

Compared to the above knitting, my rant is more of a playground whinge.

Someone has been ducking g about with the made of iPhone devils radio stack, either the Cuptino crew or the Oticon lot. My Shiny New hearing aids have made for iPhone devils radio functionality, and up to last week they fairly happily swapped between my phone and fondleslab for audio streaming g with little drama. Occasionally they needed a ‘power recycle’ to facilitate the nice playing, but mostly they worked. This week they are very reluctant to auto switch and need all sorts of fiddling. Bastards.

Recently acquired a digital audio player which I've been using to stream some music via WiFi. I'm also using Bluetooth headphones. Combining two flavours of devils radio seems to be all kinds of temperamental.
I've resultantly started re-ripping the CD collection, replacing what I'd previously managed to rip - mainly 128kB/s MP3s - with FLAC files.

It was a sad day when Apple dropped the 3.5mm headphone socket from their phones and then everyone blindly followed suit.

[rafletcher]

It was a sad day when Apple dropped the 3.5mm headphone socket from their phones and then everyone blindly followed suit.

For the time being my wife uses a lightning to 3.5mm adapter on her iPod. Works flawlessly.

[Kim]

It was a sad day when Apple dropped the 3.5mm headphone socket from their phones and then everyone blindly followed suit.

There seem to be enough people who want one that they're still alive and well in the low-mid spec Android world.  It's fairly high on my list of desirable phone features, as I want to use a set of no-nonsense in-ear headphones in an unpredictable way when travelling, and pretty much never the rest of the time.  Cynicism about devil's radios aside, I really don't want to have to charge another thing or faff about with dongles when I'm cycle-touring.

[Tim Hall]

Got home the other day to hear a rhythmic ticking. The rhythmic ticking of an external drive trying, but failing, to actually work. Arse.  There'd been a power blip, long enough to reboot the router, short enough to not affect the oven clock. And long enough to banjax this external drive. (More likely the spiky nature of the blip caused the problem).

Howsumdiver, the $Stuffs stored on it are duplicated on another external drive, hooked up to my Raspberry Pi, and are all present and correct. Hurrah after all.   

[Mr Larrington]

Oracle, you pack of scrofulous mutant baboons!  Your Java Updater has been b0rked for months if not years.  Instead of providing an elaborate workaround, why not just FIX THE FUCKING UPDATER

[Afasoas]

It's been a week of firefighting.

Latest saga... a clustered database randomly losing nodes, resulting in business essential application falling over. Turns out to be an NTP problem.

NTP clients had zero true chimers and NTP servers with offsets in seconds and jitter also in seconds for their associated NTP servers..
The NTP servers are Windows domain controllers. Virtualised. Turns out they were configured to have their time synced periodically with the hypervisor host. That explains the jitter then!

Policy had been applied to configure the PDC Emulator (root of the domain heirarchy) to synchronize time with external NTP servers. Firewall rules would not allow NTP packets to pass.
Said policy configured the source NTP servers as

Code: [Select]

Type: NoSync. WTAF.

NTP has never worked properly in this domain. It must have been like that for years. I'm really surprised this has not caused problems before. Or maybe it has and people have glossed over them.

All resolved now.
If people can't be bothered to test the infrastructure they configure, they have no business pressing buttons.

[SteveC]

Mailchimp. Will not let me save a subject line, but will then not let me send the emails without one.

ETA: about half an hour later, it behaved properly. Weird. 

[Feanor]

It's been a week of firefighting.

Latest saga... a clustered database randomly losing nodes, resulting in business essential application falling over. Turns out to be an NTP problem.

NTP clients had zero true chimers and NTP servers with offsets in seconds and jitter also in seconds for their associated NTP servers..
The NTP servers are Windows domain controllers. Virtualised. Turns out they were configured to have their time synced periodically with the hypervisor host. That explains the jitter then!

Policy had been applied to configure the PDC Emulator (root of the domain heirarchy) to synchronize time with external NTP servers. Firewall rules would not allow NTP packets to pass.
Said policy configured the source NTP servers as

Code: [Select]

Type: NoSync. WTAF.

NTP has never worked properly in this domain. It must have been like that for years. I'm really surprised this has not caused problems before. Or maybe it has and people have glossed over them.

All resolved now.
If people can't be bothered to test the infrastructure they configure, they have no business pressing buttons.

Wow.
Time synchronisation is such a basic requirement for a Windows Domain that I find that astounding.

But MS don't make it easy.
I remember when I was setting up my Homenet Domain many years ago that configuring the Windows Time Service on the PDC was not that obvious.
They had a website called 'The Cable Guy' (might still do) who did tutorials on such stuffs. And I seem to recall that I followed one of his tutorials to get it set up. It involved command-line stuffs and registry editing at that time.

So I'm now synced with an external NTP pool, and this is synced to all the Windows clients and all is good.

But you'd expect IT pros to be somewhat more on the ball than a random home user...

Is it perhaps that the initial configuration when the machines were all physical worked just fine, and then when they virtualised it they just expected everything to just continue to work as before?  The problem being introduced by the monkeys brought in to virtualise everythng and scrap the physical servers just hadn't thought everything through?

I've also seen issues with Certificate Authorities being broken by shifting to VMs without anyone considering the implications...

[Afasoas]

It's been a week of firefighting.

Latest saga... a clustered database randomly losing nodes, resulting in business essential application falling over. Turns out to be an NTP problem.

NTP clients had zero true chimers and NTP servers with offsets in seconds and jitter also in seconds for their associated NTP servers..
The NTP servers are Windows domain controllers. Virtualised. Turns out they were configured to have their time synced periodically with the hypervisor host. That explains the jitter then!

Policy had been applied to configure the PDC Emulator (root of the domain heirarchy) to synchronize time with external NTP servers. Firewall rules would not allow NTP packets to pass.
Said policy configured the source NTP servers as

Code: [Select]

Type: NoSync. WTAF.

NTP has never worked properly in this domain. It must have been like that for years. I'm really surprised this has not caused problems before. Or maybe it has and people have glossed over them.

All resolved now.
If people can't be bothered to test the infrastructure they configure, they have no business pressing buttons.

Wow.
Time synchronisation is such a basic requirement for a Windows Domain that I find that astounding.

But MS don't make it easy.
I remember when I was setting up my Homenet Domain many years ago that configuring the Windows Time Service on the PDC was not that obvious.
They had a website called 'The Cable Guy' (might still do) who did tutorials on such stuffs. And I seem to recall that I followed one of his tutorials to get it set up. It involved command-line stuffs and registry editing at that time.

So I'm now synced with an external NTP pool, and this is synced to all the Windows clients and all is good.

But you'd expect IT pros to be somewhat more on the ball than a random home user...

Is it perhaps that the initial configuration when the machines were all physical worked just fine, and then when they virtualised it they just expected everything to just continue to work as before?  The problem being introduced by the monkeys brought in to virtualise everythng and scrap the physical servers just hadn't thought everything through?

I've also seen issues with Certificate Authorities being broken by shifting to VMs without anyone considering the implications...

I was staggered when I saw the offsets and jitter ... I've never seen client associated NTP servers with simultaneously both large positive and large negative offsets.
Windows Domains are, in general, fairly tolerant to time sync discrepancies .. but I come from a background of running clustered applications with reliable time sources being essential.

I would ideally have monitoring in place to detect time drift between PDC and it's external time sources, non-PDC domain controllers and PDC, and then in-turn servers and domain controllers with appropriate alerts setup. I've even jerry rigged up monitoring of network equipment using SNMP. Sadly I don't have time to do anything but put band-aids on things.

And don't mention AD Enterprise CAs.
That was yesterday's battle.

[Kim]

I remember when I was setting up my Homenet Domain many years ago that configuring the Windows Time Service on the PDC was not that obvious.

Windows Time Service is on my list of reasons that Windows still isn't ready for the desktop.

[Afasoas]

I remember when I was setting up my Homenet Domain many years ago that configuring the Windows Time Service on the PDC was not that obvious.

Windows Time Service is on my list of reasons that Windows still isn't ready for the desktop.

Beginning with Windows 10 version 1607 and Windows Server 2016, W32Time can be configured to reach time accuracy of 1 s, 50 ms or 1 ms under certain specified operating conditions.

Courtesy of Wikipedia

W32Time is fine for most on modern versions of Windows. If accuracy is vitally important, that necessitates some gentle reconfiguration.
Of course, that's not going to be something the average user is willing/able to do.

I'll never be first in-line to advocate for anyone using a proprietary operating system and neither would I personally use Windows time sources for critical time keeping - but I'd be denying evidence I'd seen if I didn't proffer some sort of defence here. It can certainly be accurate to within 30ms as that's the threshold I'd configured for the monitoring in $oldJob and I suspect in reality that could have been a lot lower.

[Kim]

Being used the *nix ntpd, it's not so much the accuracy (though I do wonder why it has to be quite so bad) as the way it's such a pain to configure.  We shouldn't need to be scrabbling about on the command line just to configure more than one upstream server.

On top of that, not being properly compatible with everyone else's ntp implementations is just rude and annoying.

[Mr Larrington]

And another thing, Mega-Global Database Company of Austin TX, USAnia, you bunch of fucking bozos!  If I tell you to install Java in C:\Program Files\Java\jre-1.8 don't put it in C:\Program Files (x86)\Java\jre-1.8.  Because it's not the same fucking directory and putting it where you want it instead of where I told you to put it breaks Stuffs and makes me waste time fucking about with environment variables.

[Afasoas]

And another thing, Mega-Global Database Company of Austin TX, USAnia, you bunch of fucking bozos!  If I tell you to install Java in C:\Program Files\Java\jre-1.8 don't put it in C:\Program Files (x86)\Java\jre-1.8.  Because it's not the same fucking directory and putting it where you want it instead of where I told you to put it breaks Stuffs and makes me waste time fucking about with environment variables.

Sounds like you've installed the 32-bit version of Java?
IRRC the installer should just ask for ProgramFiles and get the appropriate answer depending on it's bitness.
If the installer is hard coded, it would be bizarre for the 64-bit version to use C:\Program Files (x86) ... but it's not like that hasn't been seen before.

[Mr Larrington]

Downloaded the installer from the useless pack of arsesnufflers the other day after their updater failed for the umpteenth time this year.  One would hope that they’d automatically direct one to the appropriate number of bits, but then again, perhaps I'm being overly optimistic expecting their stuff actually to behave sensibly.  I will check its bittiness after I've et me tea and kill Larry Ellison if it’s wrong.  In fact, I might just kill him anyway

In any case, if the installer asks whether I want to change the install directory and I tell it “Yes and put it HERE” is should do as it’s fucking well told

[Feanor]

I suspect that Windows itself does some redirections under the feet of the installer.
It sees a 32 bit installer and silently redirects regardless of what the installer asked for.

[Mr Larrington]

Other Stuffs which have had Program Files (x86) as the default are happy enough to go in Program Files instead.  Also, stop bringing logic into my rant

Edit: also, Java WAS previously installed in Program Files, so it worked before and now doesn’t.

[Feanor]

Other Stuffs which have had Program Files (x86) as the default are happy enough to go in Program Files instead.  Also, stop bringing logic into my rant

Edit: also, Java WAS previously installed in Program Files, so it worked before and now doesn’t.

Windows has changed over the years, for security reasons.

For example, many programs stored config information etc in their own Program Files folder.
Later versions of windows blocked programs from writing there, and these programs would have broken.
But Windows tries to fix it under the hood, by silently re-directing such writes to the hidden appdata folder, so the programs still work, believing they are writing to the Program Files folder.

In regard to the Program Files / Program Files(x86), this is a bittiness distinction.
If you are trying to force an install into the Wrong One, I can't see that really being a good idea.
It might be working for you at the moment, but just Gonnae No Dae That, Goannae No?

[Feanor]

And don't mention AD Enterprise CAs.
That was yesterday's battle.

Ah, my Homenet CA is called "Homenet Phoenix III", on account of me fucking up the original, then two subsequent Phoenix re-births...

[Afasoas]

And don't mention AD Enterprise CAs.
That was yesterday's battle.

Ah, my Homenet CA is called "Homenet Phoenix III", on account of me fucking up the original, then two subsequent Phoenix re-births...

Running a Windows Enterprise CA at home is a display of unquestionnable commitment.
I do have an OpenSSL CA knocking around but these days I understand the cool kids use a registered domain for their home networks and get a wildcard certificate issued from LetsEncrypt.

[Feanor]

This was all done as part of self-education.
I ended up knowing more about how AD domains worked than the IT staff at my old workplace!

The CA now only provides the cert for the IAS (It's not called that anymore) Radius server to dole up for WiFi clients. I use WPA2 Enterprise.
But I plan to shut that down, and revert to regular WPA authentication, to reduce hassle for visitors.

I will then shut down the domain, and the domain controllers. The only thing I need to do is make sure I can copy over the domain user files to local accounts on the client machines.
I can always just force the ACLs to Change Ownership if necessary.

But that's a bunch of work for Ron.
LateR on.

[Kim]

Okay, what has daylight saving broken this time?

I'll nominate RRDtool, which has excelled itself by handling the time change differently on different graphs with the same x-axis.

[Mr Larrington]

Notepad++, you horble bloody bast!  WTF have you done with that file I told you to save before a Windows Update-required reboot?  Bring it back immediately

(FUMMS)