On Dec 23, 2010, at 10:14 PM, Matthew Kaufman wrote: > On 12/23/2010 5:52 PM, Julian Cain wrote: >> On Dec 23, 2010, at 6:33 PM, Matthew Kaufman wrote: >> >>> On 12/23/2010 10:00 AM, Eugen Leitl wrote: >>>> On Thu, Dec 23, 2010 at 09:45:58AM -0800, Alex Pankratov wrote: >>>> >>>>> Very good question. The bonus question is if the trigger >>>>> was "external" and if it involved messing with the actual >>>>> (decrypted) Skype protocol messages. >>>> Any supernode compromises reported yet, or is this merely >>>> a DoS? >>>> >>> http://blogs.skype.com/en/2010/12/another_status_update_skype_st.html >>> (watch the video) confirms that the outage was caused by "a number of" >>> clients crashing, and that the cause of that crash was not an external >>> malicious attack. >> I've build a skype supernode probe utility and it will be run here shortly >> this evening. The intent is to determine exactly what versions of skype are >> offline by "speaking skype" and asking them for their client software >> version. > Sounds like a plan. If you have the same data from before the outage, it > would make for an interesting comparison. >> I'm starting with the supernodes that ship within the obfuscated binary >> because almost all of these crashed simultaneously from what I've seen in my >> own code logs. > Were these all/mostly up before the outage?
> 50%. I don't have an exact number. >> I don't buy what the skype CEO is saying because first it was very many >> supernodes not a few > Lots of people are saying that their Windows client crashed last night. If > you consider that almost all Internet-connected PCs run Windows it wouldn't > be surprising if most Skype-running PCs also run Windows... not to mention > the fact that the Windows version is nearly always ahead of the others in > features, so might have the latest supernode-capable code in it. The supernode code on mac and windows are very similar. The only differences are the network abstraction layers and cpu, rtsc, etc and "assembly" code. > > It only needs to be enough to make the network unstable enough that most > users think there's an outage, of course. >> also these went offline all around the same time and never came back up >> however if any are back online my utility will tell me. > Well, it is just as hard for a node that had been elected to be a supernode > as any other to get back online once a big peer-to-peer overlay goes down... > and there's the added complexity this week that (as others have pointed out > here) lots of the clients are run by people who are taking the next few days > as a holiday. > With 100,000 "known" idle supernodes waiting to pickup the slack it's hard to imagine. The election takes place very early and for the most part they only move from idle to non-dle. > Matthew Kaufman > _______________________________________________ p2p-hackers mailing list p2p-hackers@lists.zooko.com http://lists.zooko.com/mailman/listinfo/p2p-hackers
