Discussion Locked

The Gauntlet (What's missing in windows?)

0 helpful
320 replies
Closed to new repliesThis discussion is closed. It stays readable, but it takes no new replies.

This is page 4 of the discussion. Read the opening post

320 replies

by ungle

Not only are you not hosed, if you don't like thg GUI change the shell to cmd.exe and you won't use one!

The GUI is not an integral part of Windows, it's a convenience...that's why a large number of servers run happily while almost never being logged on to...

by apotheon

It's an integral part of most of what you can actually do with a Windows machine. What good is a Windows desktop without a GUI?

Besides, my point wasn't that the Windows computer is "unusable" without a GUI (though it might as well be, for most implementations), but rather that when the GUI crashes, so does everything else. Sure, you can get it running again without the GUI running, and with some work you can even get it running again with the GUI, but unlike a unixy OS, you can't just kill the problematic GUI processes, fix the problem, and restart them.

by jmgarvin

You missed the point. You should probably go back and read Apo's post.

Also, it is damn near impossible to setup MANY tools in Windows without a GUI. Try to setup Exchange without the GUI...

by ungle

OK, very nice.

Now tell me the permissions that should be granted to each key, and why the previous statement about users having a free for all on the registry was wrong...

by apotheon

When in doubt, fall back on asking obscure questions that lead off the topic at hand so that hopefully people will stop arguing against you where they're on solid ground. Confuse the topic. Great tactic.

Your previous comment about how application developers are at fault for not understanding how Windows works is just precious. Really. Think about the ramifications of that concept for a moment.

You're talking about basically saying "Well, the registry is difficult to use the way it was intended, there's no clear guidance on how to use it so that everything always works, it's fragile enough that idiots that don't know what they're doing can screw it up, and it provides a simple, centralized interface for any old security cracking malware developer to screw with system settings in some very frightening ways. Obviously, it's the developers' collective fault. Blame them."

. . . and if you think there's any kind of worthwhile privilege separation and protection in the registry, you just aren't paying attention. It is, in fact, you who fails to understand Windows, and/or what privilege separation means.

Before you start telling me that I don't understand the registry, by the way, it might help to know that I made my living for a couple years registry hacking from the command line to recover systems so screwed up that was the only way to "fix" them enough to recover persistent application state. Don't assume that just because some of us use Linux more than Windows at present we don't know anything about Windows. It was, in fact, my increasing and more in-depth knowledge of Windows that drove me into the arms of other OSes in the first place.

by jmgarvin

Ok, your inital post was that people "don't understand the registry."

Ok, I showed you that not only do I understand it, but WHY it is insecure...

Now you are throwing in strawmen...So what doesn't make sense in my post? What SPECIFIC question(s) do you have about having access to the registry that users shouldn't have?

by ungle

You can do just the same. GUI causing you a problem? Kill it, start over. DESKTOP causing you a problem? Kill explorer.exe and re start it.

Killing explorer does not hurt running applications, nor does killing applications hurt explorer.

What good is a desktop machine without a GUI, *nix OR windows? Clearly I was talking about servers.

Look, reading your posts it's clear you know a LOT about Linux (*nix?). You should assume that there are people out there who know a LOT about Windows. Maybe even more than you.

Let's learn from each other. Antagonism is boring, and not constructive.

by ungle

jmgarvin, I will not BS you.

As you mentioned, HKEY_CLASSES_ROOT is a view combining local machine settings and current user settings. Some entries are contained in HKEY_CURRENT_USER\Software\Classes and override the entries that which are help in HKEY_LOCAL_MACHINE, which contains the default settings. This, obviously, does not affect other users. Well designed separation. It means I can, for example, change file associations for user x while user y remains unaffected. Current user settings are owned by the user, and as such they have the power to change them. Don't like the settings you've got? Delete them (from current_user) with no risk to the machine, and restore the defaults.

HKEY_LOCAL_MACHINE contains the defaults, and is accessible by power users and above. There's a good reason for this - it affects everyone. If one was so foolish as to give regular user accounts power user rights they could indeed affect all on the machine, but of course we'd never do that, would we? By default, members of the local users group has read only access to the defaults. Seriously, I will not BS you. If you were to talk about 16 bit Windows (95/98) there is serious problems with the registry. I do not consider these applications to be os's and I'm sure you'd agree. They were and are crap.

Now, let's talk priviledge separation, Mr. Apotheon.

In Windows registry, I can set a security descriptor for each and every registry entry. Comparable to *nix, that would be like controlling access each separate line in your config files. *nix has terrible priviledge separation (that was obviously a joke - not a very good one)

I have not, and do not, claim that the registry is difficult to understand and/or use properly. Rather I suggest developers are lazy. My living is made teaching these developers the right way to program a Windows machine. Where they can and where they cannot write registry values, for example. What will and what will not require a user to have elevated priviledge, and how to overcome the necessity.

The registry has flaws, but I believe it's a good system overall (the biggest being there's no easy way to compress it!). It's an in-memory database I can use to store application configuration data, meaning access to it is FAST. All you have to do is spend a bit of time on the MS website finding out where is appropriate for your values!

As for users destroying machines throught the registry...you've got me on that one. If you are using NT based OS (and heaven help you if you are not) then the ONLY thing a user can do (and by user I mean someone with user piviledges only) is to screw up his own environment. If he screws it up, log him off and delete his profile. See the description about merged views for CLASSES_ROOT.

The ONLY time I've had a problem with the registry is when one of the guys who worked for me wrote a script he ran under admin credentials, and wrote a blank environment value to the session manager during an installation. This rendered the machine unbootable, but was easy to fix - but it requires an above user-level knowledge of Windows. He was nearly fired for that stunt.

If he had followed my instructions and installed his application with MSI, it would not have happened. Msi will not allow you to do that.

If either of you have SPECIFIC questions about the registry, I'd be happy to answer them. This is what I STILL do for a living. 3 years ago the company I was with did a study on TCO/ROI between Windows and Redhat for certain jobs within our environment. Windows was cheaper to buy, cheaper to get support for and cheaper to manage FOR THOSE TASKS. There were no advantages that made the extra cost of Redhat FOR THOSE TASKS justifiable. I made the decision to stay with MS...for now. I review that decision regularly for each task I manage.

by jmgarvin

I'd post my points now, but I don't have the time. Needless to day:

A) There is no clear seperation of user and kernel space in Windows. This is a huge issue that must be fixed before much more can be done.

B) If the registry worked as you described, than we wouldn't have malware issues. The biggest gripe I have is the mudging together of user "stuff." HKEY_USER is still a superset of HKEY_LOCAL_USER...this is problematic because we can have global keys for users that shouldn't be...tyically LOCAL_USER is limited in scope and USER contains the values (even those it shouldn't) for even LOCAL_USER in many cases.

C) Permissions and privelege seperation ARE an issue. It is too easy to get admin privs and too easy to just step past any barriers in your way.

D) Code can do things that it shouldn't be able to do. I can execute as admin without even being admin or asking, pretty silly if you ask me. I can write values to the registry (esp with things like .Net) and create special environments outside the scope of my user.

E) Power user isn't the problem. THERE IS NO SECURITY.

by apotheon

"[i]Kill explorer.exe and re start it.[/i]"
Big difference between a crash and a restart. Maybe you should read more carefully.

"[i]Clearly I was talking about servers.[/i]"
Clearly, you're not paying attention. Clearly, you're the only person that thinks that only servers were being discussed. Clearly, you should try having the same conversation as the people to whom you're responding.

"[i]You should assume that there are people out there who know a LOT about Windows. Maybe even more than you.[/i]"
I do assume that. I also can tell from what you've said that, at least in some areas, you're not one of them.

If you want to avoid antagonism, don't tell people they're wrong when you don't have a clue what you're talking about (or, alternatively, are just lying -- but I tend to guess you're clueless rather than malicious: Hanlon's Razor, and all that).

by apotheon

"[i]In Windows registry, I can set a security descriptor for each and every registry entry.[/i]"
That's not privilege separation. Go back to school. Privilege separation is keeping users from invading each others' privilege space. What you describe is privilege definitions, which are not separated worth a damn since privilege escalation is such an absurdly easy thing to do. This ties into the point jmgarvin brings up (repeatedly, though you don't seem to be paying attention) regarding kernel and user space, among other things.

"[i]Comparable to *nix, that would be like controlling access each separate line in your config files.[/i]"
You really need to start paying more attention. Perhaps you've heard of SELinux, TrustedBSD, SEBSD, PitBull LX, LIDS, Trusted Solaris, basic Solaris zones and containers functionality, and similar kernel-level security tools in the various unices. You want fine-grained security control? Windows isn't the place to find it.

"[i]I have not, and do not, claim that the registry is difficult to understand and/or use properly. Rather I suggest developers are lazy.[/i]"
No, you said that the developers don't (or won't) understand the registry and use it properly. Of course, if it's that big a problem trying to use it properly, that most developers just say "screw it", I'd say that's a huge problem with the registry. Good design encourages proper utilization, and protects against improper utilization. Something as fragile as the registry-based system, wherein the entire friggin' system comes crashing down around your ears because some idiot developer didn't spend the extra month learning about the registry keys he'd have to use, is a huge issue that needs to be addressed.

"[i]It's an in-memory database[/i]"
There's another problem, by the way, that creates a great number of following problems.

You'll never see configuration corruption causing problems for the whole system in unices. You'll never see a need to reboot the entire friggin' computer to take advantage of a configuration change in unices. You'll never see a large memory footprint on systemwide configuration. You'll never see a lot of problems that the RAM-based flatfile database (which is exactly what it becomes the moment all those configuration values are loaded) of the Windows Registry causes when you're dealing with a unix system. I suppose you've never heard of orthogonality of design, though.

"[i]access to it is FAST[/i]"
Sure, if you need access to [b]the whole thing all at once all the time[/b], and as long as you assume you'll never need access to the persistent configuration data while the machine's running. It's fast only for specific purposes, and those purposes are for the most part useless in a well-designed system where you're not trying to integrate the operation of every single process with that of every other simultaneously. With that sort of lack of modularity, it's no wonder Windows is a house of cards in a storm.

"[i]the ONLY thing a user can do (and by user I mean someone with user piviledges only) is to screw up his own environment.[/i]"
I guess you haven't noticed the absurd requirements for administrative privileges to perform common, everyday tasks.

"[i]Windows and Redhat[/i]"
Great. Price is one of the reasons I don't use RHEL when it's not necessary -- and so far, when it is, Fedora has sufficed.

by ungle

jmgarvin, if I came across condescending, I apologise. I trap myself in that which I hate.

I would be interested in discussing/questioning your points:

a) I've heard this before, but no one can actually demonstrate it. There are some good discussions on the Microsoft website around this topic. Having said that, I can't put my finger on one at present! It does exist in the modern Windows environment. Again, I do not include Win 95/98/ME in my discussion! That said, it is different from *nix, but hey, change is not bad in and of itself!

b) Im not sure what you're saying here. Yes, HKEY_CURRENT_USER\Software\Classes does contain entries that belong to HKEY_LOCAL_MACHINE\Software\Classes - the local user's instance of those settings. Any changes made here do not affect other users or the global settings of the machine. I'm not sure I understand your complaint - how this could directly create malware vunerabilities? The biggest vunerability sits in front of the keyboard.

Case in point: My wife is one of the most technologically impared people I know (sob), but LOVES to web surf. So I locked her machine down tight. She has never been infected by virus/malware/spyware/etc. That's not an accident, believe me :-)

c) You keep repeating that. Do you have an example? Maybe I can help you lock down your system to prevent it. Aside from bugs (yes, they are plentiful) I have never seen unintentional priviledge elevation. I would very much like to see an example.

d) Again, I'd love to see an example of this. Only thing I can think of is that MSI allows priviledge escalation during install - but you have to deliberately enable that, by default it's not enabled. I've done a lot of .NET development (C# is my current language of choice), I know of no way the .NET assists you in sidestepping security, but I'd be interested in hearing about it. Actually, I'd say that non-managed laguages ar a far bigger threat

e) OK, you're p*ssed. You're exagerating, obviously. If there was no security virus writers wouldn't need to find buffer overrun vunerabilities... In a previous life part of my job was to lock down standard builds for an international investment bank - they take security seriously. Yes, we had a couple of virus breaks, but our security was never compromised. There were two reasons: A really good firewall team, and a tight build of the desktop OS.

Please, let's call a truce. I am interested in your opinions and I'd like to review some of them. I've noticed on this forum (I am not including you here) people making sweeping statements born of ignorance with no foundation - for example "How come windows virus scanners don't start untill you log on?" - and these I have no interest in. I AM interested in genuine issues/complaints against Windows. Yes I do have a couple myself. You might be interested to know I started this discussion looking for ideas for a new adminstration product I am designing.

apotheon...as I sit in front of a Fedora machine in my test lab writing this response on Mozilla, it would seem to me this is a religeous war for you. Please read between the lines in my response, you seem like a smart fellow I don't expect to have to spell everything out. As before, if you have specific points you want addressing, please ask. Otherwise, good bye. Yes, this comment is condescending. A measured level to be comparable to your response.

by stress junkie

The best spreadsheet program that I ever saw was called Multi-Plan. It ran on C64 and Microsoft Xenix running on Tandy Model 16 computers. It had features back in 1987 that you still don't see in other spreadsheet programs today.

And the C64 graphics were way ahead of other platforms for years after the C64 was created.

And the C64 was easy to program.

And the C64 had great games. My favorite was Winter Olympics.

The C64 designers had the good sense to divide the computer functions between specialized chips. This gave the computer much more power than the specs would have suggested. ( 1MHz CPU and 2MHz tri-state data bus.) This kind of division of labor is only partially implemented in today's home computers. Imagine the potential performance increase if today's home computers were to fully implement the idea of a different card/circuit for each type of function. The CPU would be freed from network card interrupts and graphic memory updates and disk i/o interrupts. This would leave the CPU available for CPU type work such as number crunching. Wouldn't that be something?

by ungle

but due to strict process and memory space separation, a crash of an application IS the same as killing it. Explorer crashing does not kill everything else. If that's your experience, get off win98.

I have NEVER seen an application crash cause any problems on the OS itself.

Wanna talk razors? Occam's razor.

I have managed up to 25,000 high demand (traders) desktops worldwide including multi-lingual systems (Mainly east asian) without any of the problems you are describing.

Occam's razor says you are doing something wrong.

I amy be stupid, but that makes you deliberately rude, and I know which I'd rather be.

by ungle

What made you think I missed the point?

On a desktop OS, you need a gui, linux or Windows. If your desktop crashes on Linux, run startx. If your desktop crashes in Windows, run explorer. Wasn't that the point? You running apps will not die in Windows if explorer crashes. Start it up again and all the running windows are still there. Wasn't that the point? I think I got the point, but Apo didn't relise it was possible in windows.

BTW, part of what I do is creating unattended (non-interactive, or doesn't require a gui) installs. Including Exchange. I can configure all of exchange start to finish without ever loading the GUI - if I wanted to, and in my job I do, and I do :-)

Yep, there's still stuff you can't do without the GUI...but not very much any more.

by apotheon

I'd rather be "rude" (blunt) than stupid and convinced of my infallibility -- and you certainly do seem to think that, despite the fact that you're basically defending one of Microsoft's worst mistakes for security and stability purposes.

Maybe you're removed from the trenches, where one deals with a wide range of problems unrelated to single-purpose high-demand systems. Most of my clients when I was consulting had to press their PCs into a wide range of duties for each system, rather than having dedicated single-purpose workstations. Under the sort of controlled circumstances to which you seem to be referring, Windows doesn't have nearly the problems that can arise when describing the business purpose of a system is more like trying to specify the exact pattern of a twenty-yard 00 buckshot hit than a well-aimed throw at a dartboard. The more diversified the implementation, the more problems arise with closely integrated system services and applications that draw on overlapping areas of system configuration.

Maybe you've just been lucky.

Maybe you're lying.

Whatever the situation, your anecdotal argument that you "[i]have NEVER seen an application crash cause any problems on the OS itself[/i]" is patently absurd.

"[i]Occam's razor says you are doing something wrong.[/i]"
Nonsense. The fact that problems exist for tens of thousands of people the world over, all involving this one OS, and that you're the only person I've ever run across who has never had problems with applications cause problems for the OS when dealing with Windows, indicates that William of Occam would probably say Microsoft did something wrong.

Maybe the problem isn't so much that you're stupid. Rather, perhaps you think everyone else is stupid, and will eat up your BS.

by apotheon

"[i]Apo didn't relise it was possible in windows.[/i]"
I did indeed know that it is possible to restart explorer.exe in Windows. I've certainly done it enough times. That's not very useful when you get an unrecoverable BSOD, though (for instance).

by apotheon

Nice how you ignore specific points and, in return, demand specific questions. Have fun with that.

by calithor

Novell is very mature in Linux already...Novell being the forward thinking company that it is is right now the only company with a Linux solutiuon form the desktop to the enterprise... Maturity comes with eDirectory...Novell has had a directory story way longer...eDirectory manages windoiw sna Linux server better that windows and lInux themselves...
All of the very best of NetWare has been ported to Linux so you may have the best of both worlds...and if you still need windows..then it operates extremely well in a heterogenous environment...PS server down time is more costly

This discussion is closed to new replies.