What performance penalty would you tolerate to have a "safe" computing environment?

[loinburger]

Assume that you can use an operating system that monitors/sandboxes/whatever all of your processes to the point that you can be reasonably certain that one process cannot undermine the integrity of another process, that a process cannot gain access to resources to which it has not been granted permission, etc. The OS itself would be crammed with enough run-time checks to reasonably assure you that the integrity of the OS cannot be undermined. How much of a performance penalty would you tolerate in this system? 1.0 is the speed of your current system, 2.0 indicates that your MIPS (or whatever measure you prefer) has been cut in half, 10.0 indicates that your MIPS (or whatever) has been reduced by 90%, 100.0 indicates that your MIPS (or whatever) has been reduced by 99%, etc.

[Lul Thyme]

EDIT: misunderstood

[Lul Thyme]

EDIT: misunderstood

[MORON]

It depends on how hard I'm pushing the computer. (which is an function of avaible software)

Basically, I can work only so fast, and if I don't have to wait for the computer, than its fine, but if I have to, that is not good.

So my calculation would be something like
Idle Waste Time == (Recovery Waste Time * Risk of failure)

[loinburger]

Originally posted by Lul Thyme
I think there needs to be more option in the high range.

75 indicates a slowdown of 75, not a 75% reduction in performance -- a slowdown of 4 would translate to a 75% reduction in performance. I wasn't sure whether I ought to go with slowdown or % reduction, because I figured that either way somebody was going to wind up making the wrong selection. In hindsight I should have put both measures in the poll, but that would have required a lot of calculator-work on my part, and I'm very lazy.

[Lul Thyme]

Lol thats funny because I even reread your post to make sure I understood. But I read "10.0 indicates that your MIPS (or whatever) has been reduced by 90%" as "10.0 indicates that your MIPS (or whatever) has been reduced TO 90%"

In this case, I would still accept a slowdown of at least 90% or 10 if you want. Because,as I said, in this industry this represents something like 5-6 years backwards, but with the knowledge not be ever worried again by viruses, hackers some kinds of identity-theft, etc...

[Skanky Burns]

Depends on what I was using the computer for, I suppose. For gaming purposes, I can live with a bit of risk while at the same time any performance hit would severely impact on my enjoyment of the game. Running spreadsheets and other stuff where there is sensitive data (especially if I'm at work) then a bigger performance hit would be justifiable. Still, going above 2 would be rather extreme.

[Mr. Harley]

Also, one can deny the question. The old mainframes, which I learned to program Assembly Language in College on (IBM 360) did not take a much, if any, of a perfromance hit.

The bigger issue would be the "convenience" hit. A properly safe environment either is going to make Microsoft scads of money (there are no "unregistered" ActiveX applets, and MS makes money in fees for every one) while granting them even more monopoly status (Java no longer exists). Thus you will have fewer web pages that are anything other than text.

Plus of course MS will have to totally reengineer how it does things. Many of the old mainframes were utterly safe. MS may well be doing this with Longhorn. So it's a nice question, but meaningless. A safe computing environment is only gonig to take a small hit in performance - but a hideous penalty in online convenience.

[Urban Ranger]

Originally posted by Mr. Harley
A safe computing environment is only gonig to take a small hit in performance - but a hideous penalty in online convenience.

I don't think that's the case at all. I work in a Linux environment, and I don't experience any decrease in online convenience. ActiveX applets are unnecessary. They may have legit uses, but the harm is much greater than any potential gain.

[Urban Ranger]

Security is not one thing or device. It should be a major theme woven in the foundation. You can have a completely safe OS, but if somebody can have physical access to your computer or if your colleagues give out passwords to strangers on the telephone, You might as well hang it up.

A safe OS is defnitely an important factor in computer security, but it is not the only one.

[KrazyHorse]

I voted 1.25. Security is not overly important to me. I operated from 1995-2003 without even a firewall, with zero problems.

[Spiffor]

1.1.

Security is a very remote concern of mine. On my computers, I generally remove the antiviruses when they slow down the computer too much.

[Blake]

I voted 1.1, altough in practise there is no compromise. Windows (with about 1% CPU used for firewall) for unsafe computing and Linux for safe computing.

Of course I don't care about the safety of the intellectual property of soulless corporations.

[Urban Ranger]

I voted 1.5. If an OS requires more than that to lockdown things, something is wrong with its design.

Things such as buffer overruns should never happen after the Morris Worm. They are caused by laziness and sloppiness and nothing else.

[loinburger]

The problem with checking for buffer overruns is that array bounds checking imposes a significant penalty on performance. A slowdown of two is not unexpected, and you can get a slowdown of four or five on some microbenchmarks.