Klaus, Thank you for your help, if this fast behavior can be documented as such, this is good enough for me. Also, nice set of references to SAT, bookmarked for future research =) Arthur V. On Tue, 2008-09-23 at 14:49 +0200, Klaus Kaempf wrote:
- Arthur Rodrigo Sawazachi Valadares <arthursv@linux.vnet.ibm.com> [Sep 23. 2008 14:44]:
Klaus,
Thank you for the quick reply, I wasn't aware of this change in openSUSE 11.
With this new type of resolver, we can now document to our clients that package resolving shouldn't take any time at all and if it does, we can guarantee that the machine is frozen and must be rebooted?
From my current experience, I would agree.
I know this new resolver is deadlock free, but the machine might hang for any other reason, even malfunctioning hardware. Unless we can tell our clients for sure that it will NEVER take long (by long I'm thinking more then 10 seconds), I think we would still have a possible usability problem.
Well, if you have an awful lot of package repositories added and a low-memory machine, it might take more than 10 seconds due to swapping ;-)
And as a side note, I haven't been keeping up with the latest computer science theories, but as I remember, SAT is a classic NP-Complete problem which would be the opposite to the definition of fast =)
:-) It indeed is NP-complete. But the problem domain (packages) is sufficiently small to get a reasonable runtime performance.
Nice to know that someone came up with a good algorithm (approximation?) for it.
See http://del.icio.us/kkaempf/sat for a more complete list of references.
Klaus
-- To unsubscribe, e-mail: zypp-devel+unsubscribe@opensuse.org For additional commands, e-mail: zypp-devel+help@opensuse.org