Bzzt! Thanks for playing. By definition a secure hash is one where it is computationally intractable to generate data which hashes to a particular (chosen) value.
Re: SPEWS: If blocking legitimate mail servers as collateral damage is "reliable", then yes, it's very reliable. However, I don't believe that hurting even just one innocent party is worth it.
A hash of the ebuild is rsync'd WITH the ebuild from the SAME source! Simple integrity check? Yes. Proof that the Manifest, or anything in it, is the same as Gentoo intends it? Absolutely not.
I was actually thinking about this the other day, and it seems very strange that the Gentoo people haven't realized that they could enhance the security greatly by simply rsync'ing from two different servers. The ebuilds are fecthed from the first server, and the manifests/digests are fetched from another server. This would greatly diminish the risk of a rooted mirror being used to sneak in trojans, etc.
Not really. LIDS has a super-super-user mode which you can only enter through a particular executable (which no one can tamper with if LIDS is set up properly) where you can fix things if they break. However, initial setup is highly non-trivial, mainly because some important unix programs just assume that they can create/modify files in e.g./etc (passwd and mount have particularly annoying behavior in this regard). But once you're done setting everything up, you basically never need to touch the machine again.
Yes. You can lock the system down so that not even root can do anything. The concept is called Mandatory Access Control. Try googling and reading a bit about it, it's what (almost) all the big boys use for truly secure operating systems.
SRP looks very interesting too. It's a zero-knowledge based system and does not even require encryption when authenticating to be secure from capture/replay and brute forcing. It does not require a key to be stored locally at the client (you simply use a passphrase), and the server does not have enough knowledge to reconstruct the password. Furthermore, the password is never transmitted to the server.
One caveat though: You need to generate/transmit the password in some secure way (as is the case with all systems).
Encrypting/signing 1 million emails individually is much more computationally expensive than encrypting/signing 1 email. (Although it may not be enough with e.g. OpenPGP to foil spammers, it would certainly be possible to increase the computational cost of encrypting/signing to the point where it becomes impractical for spammers).
Suppose you labor extremely hard to create something, it took so much of your time, might have cost you a marriage, every single penny in your account, and someone comes and swipes it from under your feet what would you do? Without patenting there wouldn't be much you could do now could you.
As long as we are imagining things, how about this: You labor very hard (and independently!) on a graphical app only to find that a large corporation has a patent on "a method for conveying the intention for an action to occur on a graphical display" (ie. clicking your mouse). Who's fucked now?
Remember that corporations can trivially afford to patent anything which does not have prior art whereas your small inventor cannot.
In short: Read the fucking protest page and think. Please.
No. You are ignoring the fact that one can (now, and presumably in DJB's scheme) change the IP address of the server WITHOUT everyone having to update their bookmarks.
Workrave is also a good alternative. For some reason xwrits doesn't really work with my chosen window manager, Ion -- it doesn't force breaks properly. So I tried workrave, and have not looked back since.
It would be possible to write a program to repeatedly compile the code in question with different options and flags until it hits on the exact binary that was shipped.
Theoretically possible, but NOT practical. Are you aware how many options typical compilers have? Gcc has at least 100 options. That's a LOT of possible combinations.
Yes it is. A language which lets you stomp all over memory you haven't allocated or overwrite the stack contents is fundamentally broken. Simple as that.
Well, actually... If the user application is multithreaded there is a whole class of bugs (race conditions) which cannot appear on single-CPU systems simply because the code is not actually executing on two CPUs simultaneously -- these bugs would appear on SMP systems. (This is of course a bug in the application, not the system, but I though I'd mention it.)
Slashdot Mirror
Indeed. Along the same lines as "These are the addresses of abortion clinics and the doctors who perform the abortions"?
"Yeah, uh, we put a lot of innocents in jail, but on the bright side we did also put a lot of criminals in jail."
You need to come up with something better.
Here's yet another way to put it... There are two types of Computer Scientists:
Neither of which are actual Science in the classical sense. Go figure.
Bzzt! Thanks for playing. By definition a secure hash is one where it is computationally intractable to generate data which hashes to a particular (chosen) value.
Re: SPEWS: If blocking legitimate mail servers as collateral damage is "reliable", then yes, it's very reliable. However, I don't believe that hurting even just one innocent party is worth it.
I was actually thinking about this the other day, and it seems very strange that the Gentoo people haven't realized that they could enhance the security greatly by simply rsync'ing from two different servers. The ebuilds are fecthed from the first server, and the manifests/digests are fetched from another server. This would greatly diminish the risk of a rooted mirror being used to sneak in trojans, etc.
Hmm, maybe I should file a bug report...?
No, because there is less (contiguous) free space to put new files in.
It may not mean anything in Swedish, but it does mean "old" in Faroese (yeah, that's a language), and probably in Icelandic as well.
Not really. LIDS has a super-super-user mode which you can only enter through a particular executable (which no one can tamper with if LIDS is set up properly) where you can fix things if they break. However, initial setup is highly non-trivial, mainly because some important unix programs just assume that they can create/modify files in e.g. /etc (passwd and mount have particularly annoying behavior in this regard). But once you're done setting everything up, you basically never need to touch the machine again.
Yes. You can lock the system down so that not even root can do anything. The concept is called Mandatory Access Control. Try googling and reading a bit about it, it's what (almost) all the big boys use for truly secure operating systems.
SRP looks very interesting too. It's a zero-knowledge based system and does not even require encryption when authenticating to be secure from capture/replay and brute forcing. It does not require a key to be stored locally at the client (you simply use a passphrase), and the server does not have enough knowledge to reconstruct the password. Furthermore, the password is never transmitted to the server.
One caveat though: You need to generate/transmit the password in some secure way (as is the case with all systems).
We're not elitist. We're just plain superior. :)
Apparently, you are not.
Encrypting/signing 1 million emails individually is much more computationally expensive than encrypting/signing 1 email. (Although it may not be enough with e.g. OpenPGP to foil spammers, it would certainly be possible to increase the computational cost of encrypting/signing to the point where it becomes impractical for spammers).
As long as we are imagining things, how about this: You labor very hard (and independently!) on a graphical app only to find that a large corporation has a patent on "a method for conveying the intention for an action to occur on a graphical display" (ie. clicking your mouse). Who's fucked now?
Remember that corporations can trivially afford to patent anything which does not have prior art whereas your small inventor cannot.
In short: Read the fucking protest page and think. Please.
No. You are ignoring the fact that one can (now, and presumably in DJB's scheme) change the IP address of the server WITHOUT everyone having to update their bookmarks.
Workrave is also a good alternative. For some reason xwrits doesn't really work with my chosen window manager, Ion -- it doesn't force breaks properly. So I tried workrave, and have not looked back since.
I believe the poster actually meant "rediculous", as in "diculuous once more".
That statement (by Descartes?) is nonsense from any sort of logical standpoint. It should be "I think, therefore I think I am.".
How useful is that?
Are you mad? This is slashdot.
Yes it is. A language which lets you stomp all over memory you haven't allocated or overwrite the stack contents is fundamentally broken. Simple as that.
Well, actually... If the user application is multithreaded there is a whole class of bugs (race conditions) which cannot appear on single-CPU systems simply because the code is not actually executing on two CPUs simultaneously -- these bugs would appear on SMP systems. (This is of course a bug in the application, not the system, but I though I'd mention it.)
I can imagine how being required to provide formal proofs of all code would increase anyone's productivity. :)