Memo to Microsoft: Windows 10 is broken, and the fixes can't wait Windows isn't working – and Microsoft urgently needs to change how it develops the platform, and jettison three filthy practices it has acquired in recent years. In 2014 Microsoft decided it could do a better job if it discarded a lot of software testers. This bright new dawn was lauded at the time by Peter Bright at Ars …
This post has been deleted by its author
"QA still exists and is still important, but it performs end-user style 'real world' testing, not programmatic automated testing. This testing has been successful for Bing, improving the team's ability to ship changes without harming overall software quality,"
Yeah, well, Bing aside, they've got it precisely backwards. They should ship software that increases software quality without introducing changes in either UI, functionality or "features", especially when the "features" are less useful than a break-dancing poo emoji, (coming soon, exclusively for enterprise!).
Yes, I think their conclusions are on shaky ground.
"QA still exists and is still important, but it performs end-user style 'real world' testing, not programmatic automated testing. This testing has been successful for Bing, improving the team's ability to ship changes without harming overall software quality,"
An equally valid conclusion is that Bing's overall quality is unaffected by the kind of testing carried out. See turds, and the efficacy of polishing same.
According to the article they have kept QA for real-world end-user testing but dropped automated testing in favour of crowd-sourced testing... WTF?
You can't replace automated tests - unit tests and so on - with human tests because some of these things are not directly testable at user level. Surely, crowd-sourced testing should (if should is the right word) replace some of the end-user testing instead?
"This testing has been successful for Bing, improving the team's ability to ship changes without harming overall software quality,"
--Don't you need some quality to begin with in order to harm it?
But joking aside, what is Microsoft's rush? There is no additional revenue in any way I can understand by rushing to update and add 'features' to software that people have already purchased, one way or another. And the slip in quality control is likely scaring businesses and causing IT departments to take a hard look at more stable platforms, such as Linux.
Really, MS has just confirmed what its users have suspected for years--that they test in production and the world is their beta testers. And like many IT folks, I have a love/hate relationship with Windows--it keeps me employed but makes my head pound with "What were they thinking??"
yep NEVER any QA issues in the world of Linux!
Ah, bollocks, balderdash and tosh! There is no such thing as "Linux" as an operating system. You want stability? Run Debian stable. You want a decent desktop? Mint. A server? Debian. Paid for support? Dead rat/SUSE/Oracle or any number of vendors who would love to bite your arm off at the shoulder.
What you're talking about, my anonymous commentard, are distros. Some are good, some are mediocre, some are bad. The point here is you can CHOOSE, which is something MS don't want you to do with Windows. How many times have 7 users been plagued by GetWinX and trying to keep the spyware out? Choose to run Windows 7? We'll make your life difficult until you capitulate and learn to love Cortana.
When was the last time you saw a GNU/Linux distro telling you that update to Dead Rat 10 or GTFO are your only options?
well... you took the bait.... The most disappointing part is the lack of ability to focus on the issue at hand -- declining quality and QA problems in Redmond's release cycle. After having used windows since the says of Windows 286 on DOS 3.3 (remember the 32M partition limits?) one thing is clear. While there have always been patch releases at some interval (the ".1" or ".1a" releases long before the "SP1,2,3" days) windows releases have always been fairly solid. Yes exploits were found/created and the evolutionary arms race between miscreant and patch have settled into a monthly battle, but overall the windows releases were largely free of QA debacles. Through 386, 3.0, 3.1, 95 (OK, that was pretty bad), 98, ME, XP, 7, etc.. the OS and any SP's were relatively fine. With 10 and disbanding the dedicated testers in favor of a crowd-sourced QA, the wheels have all but fallen off. You can't compare the OpenSource community involvement with Linux to crowd sourced insider QA with windows. It's apples and oranges. Why? What can insiders do? At most they can complain about what doesn't work and provide diagnostic data, event logs, etc.. but have no clue as to what broke or where and have no way to communicate about the implementation details, or help with in any way beyond saying "It's broke". The crowd sourced insiders have no access to the source (and 99.9% have no clue what 'cl /nologo /W3 /Ox /Foobj/ /Febin/myexe /TC list of sources and libs' begins to describe) On the other side, the users are often the ones that develop and submit the patches and additions to whatever code or desktop is at issue. How someone running a software company would fail to recognize that fundamental distinction is simply bewildering. Giving dedicated testers the boot and relying on well intentioned, but ill equipped, insiders is a recipe for disaster. If you are going to charge for software, you have to make sure what you are selling works. If you are giving software away in exchange for help developing it -- you may expect a few more bumps along the road. And, that's fair.
Systemd!
Devuan, Debian+sysv-init, many others that I haven't tried because Devuan is familiar enough to me that I didn't see the point...
Don't conflate Lennart's pet project with Linux. As an aside, I'm wondering if the corpse of SCO isn't twitching again since IBM bought Red Hat. Big Purple.
"yep NEVER any QA issues in the world of Linux!"
Let's not forget that most Linux distros are FREEWARE for any purpose. While Windows costs individuals and companies a LOT of money. The quality gap has really narrowed between Linux and Windows. IMO most Linux distros are more stable, and have a lot less schizophrenic UI than Win10. They don't nag you idiotically, the menus make sense, and they don't collect telemetry. True, they don't have some of the Win10 features, but since said features often work rather erratically and spy on you unless you take pains to prevent it, I know what I'll keep using.
I'm not saying Linux is without flaws, but it's FREE, and very good. Try running Windows updates on a fairly up-to-date machine next to a freshly-imaged Linux machine getting updates and I'm still betting on the Linux machine finishing first and not needing a reboot.
Bing seems to work fine for me.
If I search for "Firefox", or "Google Chrome", I get as the first link:
"Promoted by Microsoft - Microsoft Edge is the faster, safer browser on Windows 10 and it is already installed on your PC"
Then the second link is the link to the relevant browser.
If you do this from a computer that isn't running Windows 10, then the first link is the link to the relevant browser.
That's the only thing Bing is used for, and it does it just fine.
Actually, "At Bing, the task of creating programmatic tests was moved onto developers, instead of dedicated testers."
The failure was *moving* programmatic tests and not *collaborating on*.
Suggests the developers had a culture of not testing themselves (In the Bing team - This may or may not be true) and that would certainly help if they started doing some basic unit/integration tests.
There are times I strongly disagree with Andrew and delight in firing torpedoes below his waterline, but I'm 100% behind him on this, Windows update is utter shit.
Even number one MS ass kisser Ed Bott says it's too much:
https://www.zdnet.com/article/opinion-two-windows-10-feature-updates-a-year-is-too-many/
I strongly agree with Andrew on his premise, not his solution. Hiring more testers now is not going to fix the problem. Microsoft has clearly demolished the corporate culture that once supported the Windows multi-million-line codebase.
In the 1980s, I visited the MS 'campus' many times. There was an electric hum in the air. A feeling like NASA mission control, of many parts working together in perfect synchronization. Over the past couple of decades, all reports indicate that this culture of precision, responsiveness and attention to detail no longer exists. This kind of collapse is self-reinforcing. MS used to be a Mecca for software geniuses; now it's just a name for ambitious execs to put on their CVs.
A corporate culture is like any fragile ecosystem: once it's lost, it would take a miracle of reverse entropy to ever see it rebuilt.
@fing0
Quote: "MS used to be a Mecca for software geniuses"
*
Err....no....that's not right. Better: "MS used to be a provider of cash for software geniuses":
- MSDOS - bought from Tim Paterson (who copied the design of CP/M)
- MS C - bought from Lattice
- MS Powerpoint - bought from Powerpoint
- MS (Visual) Foxpro - bought from Dave Fulton
- Internet Explorer - licensed from Spyglass
*
.......I could go on...........Did MS actually ORIGINATE anything at all?
Altair Basic/Basica/GW-Basic, developed by Bill Gates and Paul Allen themselves. Not much beyond that. They did make Word for Macintosh (developed originally for the Unix clone Xenix, then to DOS, and then on the Mac as first GUI version) ... but with heavy input from Apple and Steve Jobs, with things they really really did not want to do (like proportional fonts), but Apple forced them to.
Did you know that their first OS was a Unix clone called Xenix? Oh how different the world could be...
> Altair Basic/Basica/GW-Basic, developed by Bill Gates and Paul Allen themselves.
When Bill and Paul were at Harvard they were able to use the DEC computer. It is alleged that there was an open source version of BASIC that Bill was able to obtain. When they developed Altair BASIC they used a DEC-20 to cross compile to 8080 using standard Intel tools (and did not pay for the computer time they used). The maths routines needed complete rewrite which was done by Monte Davidoff.
> Did you know that their first OS was a Unix clone called Xenix?
Xenix was a real Unix version 6 or 7 licenced from AT&T and ported to 8086 by 'The Santa Cruz Operation' (SCO), a software house that specialised in porting Unix to various machines. Later this was sold to SCO becoming SCO Xenix and later updated to System3 and renamed as SCO OpenServer.
> To be fair, the design of CP/M was also ripped off - from the PDP-8 operating system OS/8. The fact that the Copy command was called PIP (Peripheral Interchange Program) is the smoking gun.
CP/M was developed on DEC machines at Intel while Gary was under contract to develop PL/M compilers. The CP/M BDOS was written in PL/M to prove that useful programs could be developed. Intel did not want CP/M as their plans for the 8080 did not include small computers and Gary was allowed to keep it for his own use.
That the utility has the same name of PIP does not indicate that anything was 'ripped off'.
> - MSDOS - bought from Tim Paterson (who copied the design of CP/M)
There are some that say that Tim copied the code of CP/M, specifically of version 1.2.
Both SCP and MS were OEM licenced for CP/M. SCP for their Zebra range, MS for the Z80 Softcard for the Apple II. Both had all the source code that DRI would supply to its OEMs. The BDOS, however, was written in PL/M and supplied as a binary because it was invarient. At the time there were 'commented decompilers' available for various software, including CP/M BDOS. There was also an Intel 8080 -> 8086 ASM translator.
SCP was developing 8086 processor boards for their Zebra machines and needed an OS to test with. It has been claimed that they decompiled the BDOS, ran it through the Intel translator and recompiled (with many fixups) to get the initial QDOS. This would have had CP/M file system which would have been needed as they would have built the system the system using CP/M, swapped the processor board and rebooted.
The MS FAT filesystem was added later from MS's 'Stand-alone BASIC'.
It is alleged that when IBM was testing PC-DOS, Gary Kildal was able to enter a particular command and get a DRI copyright message displayed. IBM settled giving Gary money, agreeing to sell CP/M-86 alongside PC-DOS and rewriting the BDOS - which became version 1.25.
The reason that it is alleged to be version 1.2 that was copied is that it had a specific bug in handling the FCBs and this bug existed in the earliest MS-DOS and PC-DOS - prior to 1.25.
SCP initially licenced Microsoft non-exclusively for 86-DOS (or SCP-DOS) running on 8086 CPUs. In theory its use on 8088 was breaking the licence. Later, MS purchased it outright and this gave SCP as many free copies of MS-DOS as required as long as they were sold with a computer. When the SCP factory burned down they started selling a V20 chip (faster 8088 clone with 8080 support) with a copy of MS-DOS. MS had to buy back the agreement for a million or so.
> - Internet Explorer - licensed from Spyglass
Spyglass wrote IE, it was not a version of Mosaic but was new code, on the basis of getting a royalty ($5?) for every copy sold. MS gave away IE and thus none were sold and no royalties were paid. Spyglass sued and eventually won a settlement of some millions, but by then the company had folded.
> they developed their first product, Microsoft BASIC, themselves,
It has been alleged that Altair BASIC was 'based on' an open source BASIC interpreter for DEC computers. As Intel 8080 development software ran on DEC it wouldn't have been too hard. Monte wrote the maths routines that were required.
"In the 1980s, I visited the MS 'campus' many times. There was an electric hum in the air."
Yes, me too. I haven't been back in years, though. I wonder if it still exudes that creepy Bill-worshipping-cult vibe like it did back then?
"MS used to be a Mecca for software geniuses"
I don't remember Microsoft ever being that, to be honest.
I also agree with the premise. Better formal testing is part of the solution, but Microsoft needs to build better quality in from the start, not try to merely test the bugs out later.
Quality is something you design in, not tack on. It's not a coat of paint.
I read the article Monday. I think the author is on the right track.
https://arstechnica.com/gadgets/2018/10/microsofts-problem-isnt-shipping-windows-updates-its-developing-them/
What do you mean by "working update system"? It works perfectly in 7, in fact, much, MUCH faster than in 10.
Well, it does once you've got that special speed-up Windows Update KB installed. Not an issue now for most, unless you've got a new Windows 7 install and you have to specifically install that one before thje rest. And even with that, some patches install out of order and once some later patch is installed, some earlier patches don't appear in the list any more.
But this is relatively recent and indicative of the same malaise that's inflicting Windows 10.
Me: "For astronomy purposes, can you set your working hours for windows update to during the night to prevent the updates interfering with your imaging?"
Response: "Err.............. astronomy............. stars................. night"
Err. Working hours supposed to prevent updates. Set them to at night, hopefully no updates at night. Updates only in daytime, when no stars visible from ground. On same page now?
Window 7 updates only worked "perfectly" if Windows 7 was installed at the correct time. I believe if you installed during early 2017 for a few month period, some of the updates broke the update system. Updates would only work again after one or two specific updates were installed, which could be installed by Windows update, but that only worked if Windows update was working, which it wasn't.
Also; yes, this, a thousand times. A service pack 2 platform update for Windows 7 that included DirectX 12 and a few other tweaks, but rolled BACK the telemetry updates would be welcome.
Then re-hire all their old QA guys and get some real testing done between each update.
I have no love for Windows 8 or 10. They just don't feel finished, and 10 has somehow become less polished over time.
My latest laptop came with Win 8.1, and once Classic Shell was installed, and I had sorted out the settings (which are all over the place) I have got it working well. It is almost (but not quite) like W7 with a service pack added.
I am very, very wary of touching anything W10 with anything shorter than a very long, preferably pointed barge pole (halberd or pole axe might do as well). Telemetry aside, I have fellow astronomers complaining that W10 will happily start an update halfway through an imaging session, wrecking data. Getting scope, camera, guide system, computer all working nicely is quite a hassle, and given the rarity of good, clear nights, the last thing you want when you have got everything working is for the OS to throw a spanner in the works. Call me old-fashioned, but I always thought two key roles of an OS were keeping programs running smoothly, and keeping data safe. Quite clearly, farming out testing to well-meaning enthusiasts is no replacement for professional software testing.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
