Vista To Be Updated Without Reboots

UltimaGuy writes "Microsoft is working on a new feature for Windows Vista, known as Restart Manager, which will update parts of the operating system or applications without having to reboot the entire machine. From the article: 'If a part of an application, or the operating system itself, needs to updated, the Installer will call the Restart Manager, which looks to see if it can clear that part of the system so that it can be updated. If it can do that, it does, and that happens without a reboot.'"

Comments

funny department

[pohl]

"from the welcome-to-the-world-of-tomorrow dept"? More like welcome to unix of yesteryear. What's with the kid that always crosses the finish line last and somehow always gets perceived as the leader?

Re:funny department

[zulux]

What's with the kid that always crosses the finish line last and somehow always gets perceived as the leader?

Maby the last kid that crossed gets all the attention because he's "special." ...

It's the same way when I show people Windows Remote Desktop.... they act like it's a big deal.

Unix had "remote destop technology" before most Unix users could afford computer monitors.

And even then, Unix was late to the party many times - I've been put in my place by old geezers when they say... "Well, my PDP-8 did that too. With punch-cards."

Re:funny department

[kuzb]

I dub the parent comment "Welcome to the flame of today". This isn't about unix, nor is it about competing with Unix. This is about fixing a long-standing user complaint. Why must there always be a comparison? Another fine example of "it doesn't matter how many problems get fixed, we'll be here to bitch about it anyway"

Re:funny department

[MSFanBoi2]

Look, we ain't talking about Unix or Linux here, we are talking about Windows.

Windows wasn't able to do this before, now it is.

What with the kid that keeps thinking that Windows and UNIX are the same and features in each have to mirror each other.

Re:funny department

[MightyMartian]

The point, other than a bit of chest beating, is to demonstrate that Microsoft loves to ooze BS about innovation, but when you look at what they do, it's been around for a long bloody time./

Re:funny department

[dslbrian]

This is about fixing a long-standing user complaint. Why must there always be a comparison?

Probably because "long-standing" doesn't really do it justice. The problem has existed in all Windows versions up to and including the ones that exist today. The comparison is because Unix and variants overcame the problem what, years? decades? ago. Imagine them saying that Vista is going to patch a 20year hole in IE, most people would compare to alternatives and say wtf took them so long...

Re:funny department

[IAmTheDave]

More to the point, it's not even that relevant. MS can install updates without reboots too, like the latest .NET Framework 1.1 update for instance, which unlike the original install of the framework, didn't require a reboot on my PC. Heck, Windows Update often runs without rebooting.

Read the fine print:

...which looks to see if it can clear that part of the system so that it can be updated. If it can do that, it does, and that happens without a reboot.

And if it CAN'T clear that part of the system? <mentok voice>REBOOT!</mentok voice>

So, I'm not sure how this is much different than before, aside from Vista will try to unload unused system dlls as well as non-system dlls?

Re:funny department

[Zathrus]

q[More like welcome to unix of yesteryear.]q

Really? Your Unix installations save the current environment when they need a reboot (for, oh say, a kernel update) and restore it completely when finished?

What this appears to do is considerably more than just try to avoid reboots (which, while improved under XP, still happen way too damn much -- for both OS and "application" patches), it actually tries to make a reboot a non-event as far as the user is concerned.

It won't work though. There's too many potential issues -- most of them security related. If you're logged in on a network it would have to remember your login info to restore that. What if you're logged into remote connections, like ssh sessions? Or ftp? Or your web banking? While these might be solvable, my guess is that solving some of them (like retaining the SSL session for the web banking) would involve some pretty massive potential security holes beyond just remembering passwords that it shouldn't.

Fixing the real issues would require a massive rewrite of the file systems, the memory manager (esp. virtual memory), and other key OS components. Unix has done it right for a long time in this regard -- delete a file in use? Sure, no problem. But it's not actually de-allocated from the FS until the current process releases it. This has its own set of issues, but they're much more managable than the ones that exist with Windows' current methods. Better yet would be inherent versioning, ala VMS's FS. Certainly disk space is cheap now compared to back then -- it's surprising that nobody's revisited this.

Re:funny department

[ZakMcCracken]

Yes, well, with today's Unix infrastructure you can do remote server access, but can you do remote *desktop* access?

Sure you can adapt server tools like X11 to do remote access, but then remote desktop involves more: can you see your local hard drive from the applications on the desktop machine that you're remoting into? Can you see your local printer, so the printouts come where you are by default and not on the printer that is connected to the remote PC? Can you hear sounds played by applications when you remote into a PC?

Same thing with fast user switching... many people said, on Linux you have long been able to open many virtual consoles under different identities... Just Ctrl+Fn between them... Ah yes but what happens when you switch consoles? Notice how it doesn't ask for your password? Which makes it applicable in many settings, contrary to the Mac or Windows versions of fast user switching which do ask for password. Feature comes in late, but right.

As to changing OS components while running... Sure, Linux has had kernel modules, FreeBSD has had a microkernel... but is there a tool to automate dependency checking, to see which services need to be shut down, to actually shut them down / unload modules, and then relaunch services?

Unix OS's "can do" a lot of things, if you accept that many of the capabilities are pushed out of the OS onto the end-user. Actually if you start thinking this way, coding pure assembly in kernel mode actually has the most features!!

Re:funny department

[NotoriousQ]

Except this already happens anyway. When a library is replaced the original is deleted, but the content still hangs around until all programs that rely on that library close, at which point the file system deletes the content.

All programs that start after library has been updated use the updated version.

The main issue is that in windows, two files can not exist under the same name (no concept of linking). (well, sort of. I am not sure if NTFS streams can be used for this.)

Re:funny department

[zulux]

but is there a tool to automate dependency checking, to see which services need to be shut down, to actually shut them down / unload modules, and then relaunch services?

No need for that in a proper system.

Let's say I want to upgrade Samba: In Unix, while the system is running, I upgrade the binaries right over the old versions. Unix is smart enough to keep the old version around and all clients that are in use continue to work just.

Any new client gets the new version.

As old sessions drop off over the next few weeks they get the new version when the reconnect. When the last old session dies, then you're fully migrated and none of the users ever noticed.

No need for "automared dependence checking" when your system was designed properly in the first place.

Re:funny department

[man_of_mr_e]

Weeks? The idea behind an in-place patch is because of some security flaw. You don't want the old version hanging around for weeks. If there's a security vulnerability, I want to restart ALL of the services that use it immediately.

Re:funny department

[pikine]

This is essentially what Unix does, but not many people understand this.

What happens when you upgrade a package?

1. Package manager removes the files of the old package.
2. Package manager unpacks files from a new package.
3. Repeat until all packages are upgraded.

When an opened file is removed from the file system, its directory entry is removed, but the inode stays on disk as long as the file stays open. So old libraries remain on disk as long as old programs are using them, which essentially creates a parallel world of different library versions---one accessible from the file system, and the other accessible from existing running programs with open file descriptors. When the program restarts, it then uses the new library.

Re:funny department

[dslbrian]

don't you think Apple (one of the, if not *the* biggest Unix distro)

Heh, ok I almost laughed at that. If you seriously think Apple is the biggest Unix distro, you have a seriously myopic view of the Unix world. Of course perhaps you were referring to *BSD (encompassing all derivatives) instead of "Apple".

Sure it might be theoretically possible in some lab setting, but the simple fact of the matter is that the Unix distribution I use every day doesn't, and so it can't be nearly as simple to accomplish as you seem to assume it is.

I'm not a kernel programmer (I have written drivers however), but I imagine it all has to do with the modularity of the kernel and subsystems. On the HPUX and Linux boxes I use daily, and also the Sun boxes I used to use, restarting core services - NFS, automount, etc. never required a reboot. Shutdown and restart the process and your good to go. Adding applications and libraries on said systems hasn't required rebooting either. I've installed major application suites (Cadence, Mentor Graphics), X, major libraries (perl, java, etc), cups print services, and so on - none of which required a reboot. Even something as invasive as swapping the video drivers can be done without a reboot. Exit X (or init 3), adjust startup script, restart X - all done.

On the flip side however is Windows, where when I installed my HP "plug-and-play" printer, it required rebooting -halfway- through the driver installation, and then again(!) after the drivers were fully installed. As near as I can tell everytime some stupid little app or game tries to install a dll it requires a reboot. The drivers for my digital camera and scanner were the same way. Why the heck do you have to reboot the whole machine to install a usb driver is beyond me. Perhaps its the peripheral manufacturer writing bad drivers, mabye its Windows, heck if I know, but thats the way it is.

Finally..

[yamcha666]

Finally! How long have we been waiting for something like this in a Windows system? Granted, UNIX/Linux/etc has had it forever and I love it that I don't have to reboot every time I install, for example, a cd burner app in Linux. So why should I have to reboot when I install something like Nero? Or anything besides the kernel or hardware for that matter?

Linux: Because rebooting is for adding hardware.

Oh, Lordy!

[overshoot]

From the description it sounds complicated as all get out. Doing dependency checks in real time while the system is running, unlinking in-use libraries, etc.

It has "fragile" written all over it.

I suppose that there are reasons why Microsoft can't just leave an inode in place after unlinking it so that processes that use it don't lose it, but is this really the best workaround they can come up with?

Re:hmmm

[dotgain]

Exactly. Will any exiting installers call the restart mis-manager? Probably not, because they don't know about it. You'll probably still get msgboxes telling you you must restart, with a single "OK" button on it, and not be able to click anywhere else. And of course there's the case of the restart mis-manager thinking it's clear to restart a service, when it isn't.

Re:Wow... Took only 30 years to catch up...

[flyinwhitey]

"Data Recovery and Application Restart

With Windows Vista, users won't have to restart their computers for most updates and application installations. Windows Vista knows which applications and services are using which files, and if a file needs to be updated, Windows Vista can coordinate saving the application's data, closing the application or stopping the service, updating the file, and automatically reopening the application or restarting the service. This capability is provided by a feature called Restart Manager.

Restart Manager works with Microsoft Update, Windows Update, Microsoft Windows Server Update Services, Microsoft Software Installer, and Microsoft Systems Management Server to detect processes that have files in use and to gracefully stop and restart services without the need to restart the entire machine. Applications that are written to take advantage of the new Restart Manager features can be restarted and restored to the same state and with the same data as before the restart."

Show me anywhere in there where MS claims this is an innovation.

Your inferiority complex is showing.

Re:I'll believe it

[Not The Real Me]

Win2K was supposed to have the restart without reboot.
WinXP was supposed to have the restart without reboot feature.

The only way I can see Vista as having this feature is if Microsoft finally includes signalling (ala Unix/Linux and most other professional operating systems).

Re:Microsoft acts like a kid.

[quakeroatz]

Not sure what stability problems you're talking about with XP. The only crashes I have are related to dying hard drives, overheating, overclocking or using an older video drivers on new 3d apps/games. All of these factors having little to do with microsoft. Windows XP is stable. Vista? We'll see....

Re:Great job, Microsoft!

[caudron]

But seriously, why do people criticize Microsoft so much for requiring occasional reboots when a much simpler application, Firefox, requires a restart every time an extension is installed.

Because when I am asked to restart firefox I don't have to send a company-wide memo that all employees accessing server X will be unable to from 12:00AM to 12:05AM---assuming no problems otherwise it's 12:00AM to when the hell ever we figure out what went wrong on reboot of a production server.

But I agree that having the restart firefox is a pain when I'm just trying to surf the web.