Of course, the most interesting thing would be if the author was willing to release the changes he did to Serapis back to the community (I have to disagree with some of his opinions regarding my design, though not enough to argue about it). I believe I asked about this before (I even offered cvs access), however, and got no reply of such intentions. It would be importantly not only because it would allow further building on Shu Yan's well done work, but also to evaluate the exact nature of the results and possible errors in the simulator code (we all fuck up sometimes, and I am concerned about a couple of the errors in Serapis 0.1 that I tried to warn the author about in December). Were that my game, I would think that Scott and I would have just cause to be upset that we are not even credited for, or given acknowledgement of, writing the simulator on which the papers simulator was based. As it is however, I am happy that my code was useful to somebody. It is a little upsetting that thinly veiled bitterness and accusations toward us shine through in text when I honestly tried to my best to thoroughly answer every question the author posed (check the archives from December last year if you don't believe me). On Mon, Jul 02, 2001 at 11:38:54PM -0700, Ian Clarke wrote: > Extremely interesting. I don't find the 20% figure that surprising, > however it will be interesting to see how proposed improvements (such as > making caching probabilistic, with a decreasing probability of caching > as the DataReply gets further from the data's source and closer to the > requesting node) will affect document longevity. > > Kind regards, > > Ian. > > On Tue, Jul 03, 2001 at 05:47:35AM +0100, Shu Yan Chan wrote: > > Long time ago Ian has purposed a research project. At that time, my own > > research project has already begun and was too late to incorporate Ian's > > suggestions into my project. However, my project does in some sense measure > > the longevity of a document in Freenet (with requests and re-insertions). > > My dissertation are available at: > > http://www.sysoft.fsnet.co.uk/join.pdf > > I am not sure whether the results presented are inline with what one would > > have expected from the Freenet's principals and routing mechanism. However, > > I was quite shocked while I was doing the simulations. Basically, I found > > that Freenet would only be able to utilise it's storage capacity at around > > 20%, e.g. if all users have altogether shared 100GB of disk space to > > Freenet, then there could only be around 20GB of documents stored in the > > Freenet. > > I originally tried to tidy up the dissertation before making it publicly > > available, unfortunately I can never managed to find the time to do so. I > > guess all Freenet developers could go straight to chapter 4 of the > > dissertation. For completeness, I am also uploading the simulator I used > > for the simulations: > > http://www.sysoft.fsnet.co.uk/SYSerapis.zip > > (I did not have time to tidy up the source code, so it would probably look > > very messy and inefficient at the moment.) > > And also because of the constraint in time, I did not managed to simulate > > Freenets with a much larger limit of the maximum number of dataStore entries > > (which would be closer to real world usage) > > > > Any comment should be directed to me, not my supervisor nor my department, > > since I am just a student, not a memeber of staff there, and am solely > > responsible for the dissertation/simulations. > > > > Thanks. > > > > Yours, > > Shu Yan Chan > > [EMAIL PROTECTED] > > P.S. Please cc a copy of your reponses to my e-mail address, as I would not > > be able to check the mailing list in the coming month. > > > > > > ----- Original Message ----- > > From: "Ian Clarke" <[EMAIL PROTECTED]> > > To: <[EMAIL PROTECTED]>; <[EMAIL PROTECTED]>; > > <[EMAIL PROTECTED]> > > Sent: Sunday, April 15, 2001 8:53 PM > > Subject: [freenet-tech] Looking for a research project? > > > > >One important issue is that of document longevity in Freenet. The > > >current approach basically removes documents on the basis of which was > > >least recently accessed, although a bias against large files was > > >introducted in version 0.3. > > > > >So one very important question is how long does a document last in > > >Freenet (assuming no requests for that document), and how is this > > >affected by differences in the size of the document and the HTL with > > >which it was inserted > > > > > > _______________________________________________ > > freenet-tech mailing list > > [EMAIL PROTECTED] > > http://lists.freenetproject.org/mailman/listinfo/tech -- 'DeCSS would be fine. Where is it?' 'Here,' Montag touched his head. 'Ah,' Granger smiled and nodded. Oskar Sandberg [EMAIL PROTECTED] _______________________________________________ Devl mailing list [EMAIL PROTECTED] http://lists.freenetproject.org/mailman/listinfo/devl
