(heh, this is a fun discussion given how little data we have to work on.) The problem sounds (to me) like one that required neither client nor server upgrades: rather, the p2p algorithm got into a state that just took time to work itself out. (This is something we've suffered on many occasions; it's nothing to be ashamed of.)
Again, the central resources we're talking about are incredibly basic: there are many examples of IM systems much larger than Skype that handle this problem just fine centrally. Even if all 9M users decided to relogin all at once, the backload of this shouldn't take *2 days* to work out. If there's an unusually expensive centralized operation that's causing the problem, it must be something unique to Skype, meaning something due to their p2p system. Which brings me back to my question: what is the supposed value of using p2p (for anything other than the VoIP and relay service), and is it working? -david > -----Original Message----- > From: [EMAIL PROTECTED] [mailto:p2p-hackers- > [EMAIL PROTECTED] On Behalf Of Alen Peacock > Sent: Tuesday, August 21, 2007 7:56 PM > To: theory and practice of decentralized computer networks > Subject: Re: [p2p-hackers] what really happened to Skype? > > On 8/21/07, Alexander Pevzner <[EMAIL PROTECTED]> wrote: > > > > 2) The situation was fixed at the server side. Client upgrade was not > > required > > This is a very interesting piece of information, and if true, points > not to the fallibility of the decentralized portion of Skype, but once > again to the fallibility of centralized Skype-controlled resources. > > I am very curious about how this fits in with the "p2p algorithm" > problem that Skype is pointing at -- if no client upgrade was needed, > and since supernodes are clients, this seems to once again indicate > centralized bottleneck rosources. I suppose that "classical" p2p > includes centralized resources (as in napster), and so code running on > these centralized components could be considered part of the p2p > algorithm... or is that too much of a stretch? > > Very informative post, Alexander -- thanks. > > Alen > http://flud.org/ > _______________________________________________ > p2p-hackers mailing list > [email protected] > http://lists.zooko.com/mailman/listinfo/p2p-hackers _______________________________________________ p2p-hackers mailing list [email protected] http://lists.zooko.com/mailman/listinfo/p2p-hackers
