On Thu, 16 Aug 2007 22:42:42 +0200, [email protected] said: > i trust you kernel developers know what you are doing, but if scares me > a little bit, that some integral and living part like O(1) being ripped off > and being replaced by something new. There's something even scarier - O(1) being taken out and replaced by multiple new things, none of which ever gets fully tested because if one misbehaves, the user says "so what" and tries another one without bothering to file a bug report... Been there, done that - I found a timing hole in the pluggable I/O schedulers that probably nobody had managed to hit because there were 3 or 4 choices available, and they just set a default they liked. Turned out there was a race condition if I/O happened while schedulers were being swapped.
Attachment:
pgpIRmaw0g7W4.pgp
Description: PGP signature
- References:
- Prev by Date: RE: [PATCH 0/24] make atomic_read() behave consistently across all architectures
- Next by Date: Re: [ofa-general] Re: [PATCH RFC] RDMA/CMA: Allocate PS_TCP ports from the host TCP port space.
- Previous by thread: Re: Re: [ANNOUNCE][RFC] PlugSched-6.5.1 for 2.6.22
- Next by thread: [PATCH RFC] try_module_get usage
- Index(es):
 |