| 1 |
Peter wrote: |
| 2 |
> Maybe I just don't :( |
| 3 |
>> e17 doesnt break the whole system |
| 4 |
> |
| 5 |
> Any alpha software can. Read the warning label. e17 has caused me to hit |
| 6 |
> the big red switch on several occasions. |
| 7 |
> |
| 8 |
|
| 9 |
Then you have a bug in your kernel (or possibly your video driver). It |
| 10 |
should be impossible for any user program to take down the system. |
| 11 |
|
| 12 |
>> wine doesnt break the whole system |
| 13 |
>> |
| 14 |
> It can trash lots of files. Seen it. Been there. Oh, and one other little |
| 15 |
> nasty about wine, you can get hit with a virus, and depending on how your |
| 16 |
> local drives are configured, this can be a bad thing. |
| 17 |
|
| 18 |
You can be hit by a virus without wine too. Just because there are |
| 19 |
currently no viruses for Linux doesn't mean that they are impossible. |
| 20 |
Also the damage a virus in wine can do is limited to the user that you |
| 21 |
are running wine as, unless of course you are stupid enough to run wine |
| 22 |
as root. |
| 23 |
|
| 24 |
|
| 25 |
> Any "good" kernel, improperly configured or used can do the same. "Why |
| 26 |
> doesn't usb work? Why doesn't nvidia work? Why doesn't ssh-fuse not |
| 27 |
> compile? Why can't I access my home filesystem?" etc. How many users |
| 28 |
> installed hardened or sellinux and went "Oh sh*t" |
| 29 |
|
| 30 |
Those are known conditions. Gentoo developers can debug them. A subtle |
| 31 |
failure caused by a bug in the kernel can be VERY hard to debug. Take |
| 32 |
for example, a few years ago there was a kernel bug (in vanilla) that |
| 33 |
caused perl to fail compiling PORTAGE_TMPDIR was on an ext3 filesystem. |
| 34 |
That sort of bug would be almost impossible to track down if it was |
| 35 |
caused by some random, untested, duct-tape patchset that the user had |
| 36 |
installed. |
| 37 |
|
| 38 |
> |
| 39 |
> Listen, I'm not going to prolong this. My point was and IS that sources |
| 40 |
> are just that. They are not applications. They must be configured |
| 41 |
> correctly to run. If you're going to promote and publish -mm and -ck, then |
| 42 |
> you can't rightly call a source based on -ck a "pos kernel." Maybe by your |
| 43 |
> standards, but not by mine, or the others who follow this particular |
| 44 |
> thread on bz and the forums. |
| 45 |
|
| 46 |
-mm and -ck are both produced by extremely competent, and knowledgeable |
| 47 |
kernel hackers. -nitro or whatever it's called these days is just a |
| 48 |
bunch of random patches thrown together by someone who knows how to |
| 49 |
munge patches without knowing their actual effect. It's actually not |
| 50 |
terribly difficult to munge patches in code that you are unfamiliar with |
| 51 |
(I have done it many times myself), but doing this in a kernel can |
| 52 |
subtle cause breakages all over the system. |
| 53 |
|
| 54 |
> There's lots of evil out there, but kernel sources are no worse than a pos |
| 55 |
> application or alpha software. |
| 56 |
|
| 57 |
Kernel sources can cause all sorts of breakages that normal software |
| 58 |
can't. For example, there is no way for a random app to cause certain |
| 59 |
strings of data to fail to be written to disk, or off-by-one errors |
| 60 |
reading certain random files. Both of these are possible, along with |
| 61 |
many other hard to detect and diagnose breakages in the kernel. The |
| 62 |
kernel is _not_ just another program, its the main arbiter of system |
| 63 |
resources and as such can break the system in ways that even the most |
| 64 |
malicious user programs can only dream of. |
| 65 |
-- |
| 66 |
gentoo-dev@g.o mailing list |
Archives