Microsoft to Patch Problem Patch

slowroller writes to mention an eWeek article about a new patch to fix issues raised in their most recent release. From the article: "The company's plan is to target the rerelease only to Windows users who are affected. In a blog entry, Toulouse said the company's patch deployment technologies will have "detection logic" built into them to only offer the revised update to customers who don't have MS06-015 or are having the problem. The glitches, which Microsoft claims affect only a tiny fraction of the 120 million installations of the patch, stem from a new binary called VERCLSID.EXE that validates shell extensions before they are instantiated by the Windows Shell or Windows Explorer. On systems running Hewlett-Packard's Share-to-Web software, Sunbelt's Kerio Personal Firewall and some NVIDIA Drivers, users complained that the new binary stopped responding."

Millions of different system configurations.

[headkase]

Everyone complains that Microsoft does not release their patches fast enough or that they don't do adequate testing. They can't win either way.

Here is the problem

[dick pubes]

The big problem when they do this is compatibility testing. I work at numerous companies where we need to read through each patch to see what they 'fix'. Now when Microsoft does this we will just have to guess what they might break in a legacy application deployed across the world.

Re:Millions of different system configurations.

[MustardMan]

No - Microsoft doesn't release patches fast enough and they don't do adequate testing. They don't win on either count.

Heh - "tiny" fraction could still be "lots"

[NotQuiteReal]

Many product vendors would love to have a tiny fraction of the 120 million installations - it would be more than their entire market!

I know this is not a popular opinion here, but MSFT really does have a tough job, if you are objective about it, from an engineering point of view.

Re:Millions of different system configurations.

[aussersterne]

Microsoft is a multibillion dollar corporation stuffed full of multibillion dollar men. They have a monopoly on the marketplace, power half of the world, and want to power the rest.

They can, will, and had better do both:

- Release patches quickly
- Release patches with adequate testing

If they don't, they should be punished.

Re:Heh - "tiny" fraction could still be "lots"

[mcrbids]

I know this is not a popular opinion here, but MSFT really does have a tough job, if you are objective about it, from an engineering point of view.

Hear here!

I agree 100%!

As a software engineer of a rapidly growing company, it's amazing to me how much higher the standard of testing and accountability has to be with each major product release. Our company has been growing exponentially, at least 2x annually. Just a year or two ago, a bug meant a few phone calls, but in the last year or so, it's gotten to where a single bug (even a minor one) can easily swamp our telephones!

The first release was like, a proof of concept more than not. It wasn't even feature complete at release - we relied on an update mechanism built in at the last minute to cover for the fact that not all the features were completed!

Not many phone calls from that issue, I might add. But, in the last year or two, a single bug affecting a relatively small percentage of our users still loads us down with dozens of issues ticketed in a single morning.

Ugh!

Since our deliverable is web-based, fixing a bug is still very fast, but we're working furiously to improve quality control testing prior to release. I can only imagine what a company with the market size of Microsoft has to deal with - when the vast majority of computing resources are in your hands, the task of dealing with bugs and updates must be simply gargantuan.

How do they do it with such a shoddy codebase?

Re:Heh - "tiny" fraction could still be "lots"

[cyber-vandal]

Considering that position was obtained and has been maintained illegally, I have zero sympathy.

Re:URL For Patch

[MobileTatsu-NJG]

"Click here for the patch."

GRRR they didn't finish testing this patch, either! Office looks funny and none of my games work!

Re:Millions of different system configurations.

[Anonymous Coward]

Yes - they should be punished - by users not buying their support contracts for following years.

Re:Affected

[baadger]

Oooo ooo I want to slam HP too.

The HP 'drivers' for my all-in-one machine come in at 180 megabytes! The interface is sheer bloat, it installs a handful of totally unnecessary (Disabling them has little consequence) services and startup processes, and there is still no x64 driver!

The HP sponsored linux drivers (HPLIP) work well on Linux 64, and it is nice to see Linux up on Windows for once in terms of hardware support.

That felt good.

Re:Apple users are nervous about updates

[ciroknight]

Hell, at least Apple machines ask you if you want to update; this latest Windows XP patch was pushed to my singular Windows XP without me even knowing about it, installed itself, and rebooted itself. It could have at least asked me if I wanted it or told me what the update was even for...

Re:Millions of different system configurations.

[Tim C]

They can, will, and had better do both:

- Release patches quickly
- Release patches with adequate testing

You do realise that some things simply take a certain amount of time and no matter how much money or how many people you throw at the problem they will not get done any quicker, don't you?

You also realise that the reason that MS release patches on a monthly schedule is that the corporate IT world demanded it, don't you?

What you are asking for, in effet, is that they a) solve problems in a certain amount of time regardless of how long it actually requires, b) do so without affecting quality and c) go against the express wishes of a large proportion of their customers.

Now, I'm not saying that they're perfect by any means, and I accept that I'm probably lucky in that I've used half a dozen machines over the last few years running Windows 2k and XP and have suffered no problems that weren't entirely hardware related, but from where I'm sat they're doing an ok job.

Re:Apple users are nervous about updates

[fact0r]

Apple updates do not have an uninstall feature. Almost every windows update does.

Mac users should be much more wary of updates for that reason alone.

Apple also is a lot less interested in enterprise customers than Microsoft. Enterprise customers are the ones that demand extensive testing and will seriously crack the shits if some funny legacy application that is absolutely critical for their business fails to run following an update.

Apple isn't too fussed by backwards compatability either. So certainly an OS upgrade (10.3 -> 10.4) is expected to break things on OS X. Pre-Vista Microsoft pretty much guaranteed that if it worked on the old version it would work on the new version of Windows.

Re:Millions of different system configurations.

[Splab]

You do realise that some things simply take a certain amount of time and no matter how much money or how many people you throw at the problem they will not get done any quicker, don't you?

If only people would realize that, especially managers. "Ohh so you need x hours to do that? Well I'll just go call this helper for y hours, then you only need x-y hours, so we'll ship on friday"... Glad I'm not doing that anymore. Incidently, we did have a few issues with the patch, but what it revealed for us isn't that there might be a problem with MS patches, but that theres a big problem with testing at our facility before rolling out patches.

MS might screw up, but it's our job to make sure that what they give us works before we roll it out.

Re:Heh - "tiny" fraction could still be "lots"

[Jeremi]

What aspects in particular are shoddy?

I haven't seen the codebase, but from using the Win32 API a bit, I noticed the following:

1. "Fill in a struct and pass it to the function" interfaces, which are very error prone (forget to fill in a field? Oops, now you have a program that works 80% of the time and does something weird the other 20%, due to uninitialized memory reads)
2. Hungarian notation used everywhere, making things hard to read
3. Unnecessary obfuscation of types (e.g. DWORD instead of long or int32)
4. Focus on backwards compatibility to the point where there are often five or six APIs for every function (granted there are some valid business reasons for doing that, but it still makes for an extremely messy and hard-to-validate interface)
5. Functions that are broken, and instead of fixing them, Microsoft simply publishes a http://support.microsoft.com/default.aspx?scid=kb; en-us;274323 [slashdot.org], forcing every Win32 developer in the known universe to have to hack up their code with an ugly band-aid instead
6. A tendency to create Yet Another New API for everything, instead of re-using existing interfaces. For example: you have a program that communicates over a TCP stream, and you want to make it communicate the same data over a serial port instead: Under MacOS/X or Linux, this is trivial: just pass in the file descriptor to the serial device instead of to the TCP socket, and you're done. Under Win32, you'll have to completely redesign your program with a custom event loop, because there is no way to select() on a HANDLE.

Anyway, those are my observations... hopefully things are better in .net land or whatever the new thing is these days.