More Windows Update WTF
-
Windows is frustratingly inscrutable.
With XP, I was able to connect to the user SMB share on my Mac (Tiger), with no password.
Windows 8 won't let me do that. net use helpfully reports, "System error 1326 has occurred." The problem is, I've replaced my PC, not the Mac (I've had this Mac for five years), so it would appear that Windows 8 itself rejects blank passwords for outgoing connections, and blames the problem on the user and/or remote computer, when a local policy is getting involved and interfering. There's a group policy for allowing blank passwords for incoming connections (which also affects runas for local accounts) but no evidence anywhere in the world that Windows won't allow a blank password for outbound connections to other computers.
Too often in Windows, you waste a load of time trying to fix some inscrutable problem, and end up feeling utterly defeated once again by the computer, the weird rules it invents and won't tell you, and the near-zero information available about pretty much anything. Then again, I've had Linux ignore attempts to remove a file that was in use (no error message from rm, just silence), yet lsof lied to me that the file was not in use. Besides, doesn't rm just unlink a directory entry? Surely unlink works regardless of whether a file is open or not?
My all-time favourite, though, is this error from MMC:
Every single suggestion in that dialog box is a) wrong, and b) trivially verifiable by MMC, i.e. someone seriously thought it made sense to suggest that the user waste their time doing the computer's work for it. It was a widespread bug where the cause was never identified. The .msc file is intact (I tried copying one over from a PC where Defrag still worked), and you can add the Defrag snap-in to a brand new MMC console and it just works. It's as though there's a killbit for that specific .msc file itself (you couldn't connect into SBS 2011 Remote Web App RDP using a Windows XP PC when SBS 2011 first came out, as MS put out a killbit (or something related) for their own RDP ActiveX control, which was particularly dick move — no idea if that's been fixed since).
-
@DrakeSmith said:
One of my favorite quotes - "Those that think Linux is free must not value their time."
This old chestnut irks me.
I do value my time. I value it enough to want to spend the vast bulk of it doing things that are enjoyable.
The total amount of time I spend dealing with problems is about the same for proprietary vs. free software. So why would I pay to buy into a proprietary software ecosystem in order to set myself up for a relentless round of tedious maintenance, vandalism repair and workarounds for vendor lock-in, when I can get an endless source of interesting creative challenges for free?
Not just dissing Windows here, either. I loathe iPads. The school has 70 of the little bastards and they are just hideous things to work with. Integrating iPads into a Windows-based school infrastructure is not easy and not fun. The limited enjoyment I get from that aspect of my job lies in feeling like I've managed to give the finger to one of our totalitarian corporate overlords by making an Apple product do something that Apple says I need to spend tens of thousands of dollars on duplicating existing infrastructure with their gear to achieve. Android is slightly less annoying in that regard, though Google is learning from Apple at a fair old rate.
I'm much happier when I can see under the hood and look at what the bits are doing, and I love having the freedom to tweak anything to make it do something a little closer to what I want. I mean it's lovely when I don't even need to, and Debian is getting continuously better in that regard; but when things go wrong I love being able to Alt-F1 out to a text console and have a poke around to find out why.
-
@boomzilla said:
I am not fucking kidding you even a little. I should probably mention that I don't normally use these machines all the time, so I typically fire them up explicitly to do updates, so the suckage that is Windows Update isn't able to hide from me in the background. It downloads in a reasonable time, and then sits there for multiple minutes on each goddamn update. Then I have to watch it do its thing while it shuts down and then while it boots up.
If you're running any antivirus, including Microsofts, you might find out that things zip along nicely once you disable all realtime protection for the duration of the updates. I've found it make big difference on both Windows XP and Windows 7.Fuck you Microsoft. Your updater sucks ass.
-
@flabdablet said:
Android is slightly less annoying in that regard, though Google is learning from Apple at a fair old rate.
The key difference from our perspective is that the Apple stuff seems a bit better integrated, and the Android systems are way easier to put your own code on; work of seconds for any conventional app. Unless you're getting the droids from a carrier who thinks that locking things down extra-hard is double-plus fun, but why would you do that?
-
People who find themselves trying to fix this issue over the phone for customers and/or aged relatives might find these URL-shortened links useful.
-
@flabdablet said:
Been getting keytop-shaped dents in my face from this issue for a while now:@ars technica said:
Windows patches are mostly cumulative. On a fresh install of Windows XP, you don't need to install all of the dozens of Internet Explorer 6 patches sequentially; you can generally just install the latest patch, and it will include all the historic fixes because it supersedes the historic patches that introduced those fixes.
Unfortunately, the Windows Update client components used an algorithm with exponential scaling when processing these lists. Each additional superseded patch would double the time taken to process the list. With the operating system now very old, those lists have grown long, sometimes to 40 or more items. On a new machine, that processing appeared to be almost instantaneous. It is now very slow.
They ain't kidding, and the result is that older XP boxes just sit there with svchost.exe consuming 100% CPU, apparently indefinitely.
-
Doesn't look like they actually fixed their brain-dead exponential algorithm; from the description, all they've done is "depreciated" a bunch of old IE patches. Which presumably means that if you happen to have a system that's been updated with one of the "depreciated" patches, the latest patches won't supersede it properly.
If nobody's actually prepared to get into the guts of Windows Update and actually fix it, that probably means they don't have anybody on staff who actually understands how the hell it works any more. Having read the odd WU log file I can't say I'm surprised - I bet it's really enterprisey.
-
@flabdablet said:
Doesn't look like they actually fixed their brain-dead exponential algorithm; from the description, all they've done is "depreciated" a bunch of old IE patches. Which presumably means that if you happen to have a system that's been updated with one of the "depreciated" patches, the latest patches won't supersede it properly.
If nobody's actually prepared to get into the guts of Windows Update and actually fix it, that probably means they don't have anybody on staff who actually understands how the hell it works any more. Having read the odd WU log file I can't say I'm surprised - I bet it's really enterprisey.
So they improved performance by 128¹×. Too bad O(2ⁿ/128) is the same as O(2ⁿ)...
¹ number pulled out of my butt
-
@flabdablet said:
If nobody's actually prepared to get into the guts of Windows Update and actually fix it, that probably means they don't have anybody on staff who actually understands how the hell it works any more. Having read the odd WU log file I can't say I'm surprised - I bet it's really enterprisey.
I think it's all based on CIM, and that's so thoroughly enterprisey that even most of the people I've talked to at IBM found it daunting. (It's definition language has all the complexity of UML and OWL, and the models built with it are by anal-rententive twerps who need you to define everything to slightest degree of variation. I don't want to work with it ever again.)
-
@dkf said:
I think it's all based on CIM
CIM - that's the thing where Microsoft looked at SNMP and reacted with "ASN.1? Hah! Not painful enough!" - yes?
-
@Ben L. said:
So they improved performance by 128¹×. Too bad O(2ⁿ/128) is the same as O(2ⁿ)...
¹ number pulled out of my butt
Yeah, when something normally takes an hour, a 128x speed improvement is absolutely pointless.
-
@Salamander said:
@Ben L. said:
So they improved performance by 128¹×. Too bad O(2ⁿ/128) is the same as O(2ⁿ)...
¹ number pulled out of my butt
Yeah, when something normally takes an hour, a 128x speed improvement is absolutely pointless.
Limits are hard. Big-O notation is abused and misunderstood so much. Not nearly as much as statistical "P-Values," but a lot.
-
@boomzilla said:
Limits are hard. Big-O notation is abused and misunderstood so much. Not nearly as much as statistical "P-Values," but a lot.
The places where the differences tend to matter are in fairly esoteric spots like advanced multiplication algorithms. For mundane stuff like sorting and searching, getting in the right Big-O class is the first thing to check for (given the particulars of the situation, of course).
-
@dkf said:
@boomzilla said:
Limits are hard. Big-O notation is abused and misunderstood so much. Not nearly as much as statistical "P-Values," but a lot.
The places where the differences tend to matter are in fairly esoteric spots like advanced multiplication algorithms. For mundane stuff like sorting and searching, getting in the right Big-O class is the first thing to check for (given the particulars of the situation, of course).First thing to check is the profiler. A shitty (but correct) algorithm that is responsible for a fraction of a percent of run time isn't worth the time to improve. The Big-O analysis is (sometimes) useful in guiding you toward a better solution.
-
@boomzilla said:
First thing to check is the profiler.
True, but with decent library implementations of many algorithms, there's little excuse for using a shitty one in the first place. It usually just means that you're inventing your own code when you shouldn't!
-
@boomzilla said:
@dkf said:
@boomzilla said:
Limits are hard. Big-O notation is abused and misunderstood so much. Not nearly as much as statistical "P-Values," but a lot.
The places where the differences tend to matter are in fairly esoteric spots like advanced multiplication algorithms. For mundane stuff like sorting and searching, getting in the right Big-O class is the first thing to check for (given the particulars of the situation, of course).First thing to check is the profiler. A shitty (but correct) algorithm that is responsible for a fraction of a percent of run time isn't worth the time to improve. The Big-O analysis is (sometimes) useful in guiding you toward a better solution.
Big-O is more useful when you know the worst-, best-, and average-case size of n (ballpark). This is easier at higher levels because logic is more domain-specific; a library usually should be more pessimistic about workload size. For small n, an exponential algorithm may even out-perform linear or even constant time; but more commonly it just doesn't matter and you should focus on clarity and maintainability.
-
Big O is mostly useful when a giant robot threatens Paradigm City and you need a hydraulically-operated fist to destroy it.
-
@Salamander said:
@Ben L. said:
7 patches later, it will become a 1× speed improvement.So they improved performance by 128¹×. Too bad O(2ⁿ/128) is the same as O(2ⁿ)...
¹ number pulled out of my butt
Yeah, when something normally takes an hour, a 128x speed improvement is absolutely pointless.
-
@flabdablet said:
CIM - that's the thing where Microsoft looked at SNMP and reacted with "ASN.1? Hah! Not painful enough!" - yes?
Possibly. I never grokked the transfer encodings for it enough to make that call. Just the way the data is arranged is enough to make your frontal lobes start looking for a length of rope and a lamppost.
