-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Monday 22 April 2002 13:31, Aaron Ingebrigtsen wrote: > From: Timm Murray <hardburn at runbox.com> > > >Reply-To: chat at freenetproject.org > >To: chat at freenetproject.org > >Subject: Re: [freenet-chat] distributed networking Hardware > >Date: Sat, 20 Apr 2002 23:02:10 -0500 > > > >Well, collisions are not the sole source of network problems, but they are > >often a major one. > > Well, if I was able to, I would do everything in my power to reduce data > collisions. I would really also like to reduce centralization as much as > possible. If just one hub goes out, a whole section of the network goes > down. If you are useing one central switching machine for a large group of > machines, that is a weak point. > > >For starters, Token Ring's maximum speed is 16 Mbps, compared to 100 Mbps > > Couldn't someone figure out a way to improve that speed?
Probably, but nobody seems to care anymore. Even IBM doesn't care, and they invented the fool thing. > > >being used in most Ethernet networks (and Ethernet has been pushed even > >higher then that). > > > >Token Ring's one redeaming quality is that collisions don't happen. Think > >of > >some of the old Native American tribes having a meeting, and they pass > >around > >a big stick or some other emblem. Only the person who has the big stick > > is allowed to talk. > > That is kind of the way internal hardware on a PC is set up. Each device > gets to talk to the processor, but the processor can't handle haveing ALL > the devices talking at the same time, so they all have IRQs, or priority > levels, or whatever you want the think of it as. If a device with IRQ 1 > wants to talk, the processor stops what it is doing and listens. If any > other devices also want to talk, they have to wait for the more important > devices to finish first. But it sounds to me like Token ring doesn't care > about importance or priority, it just passes permision around and around, > again and again, the way a Monitor passes a beam across the screen again > and again. :( That doesn't sound too efficient to me. A CPU can prioritize because it CAN prioritize. It would take quite a bit more for that to happen on a network. > > >Likewise, Token Ring networks pass around a "token", and only the machine > >that > >has the token is allowed to talk. In contrast, Ethernet's meathod is to > >stick everyone in a room and everybody shouts at once (ignoring the > > effects of microsegmentation, of course). > > This also seems like a terrible way to handle things. I would much prefer > direct routing to and from individual machines, instead of broadcasting or > Carousel networking. "Routing" only happens at the next layer up from here. The purpase of the bottom two layers (physical and data link) is to provide a physical means to move the bits through and to have some way of pushing data to a specific machine. Routing is moving the bits between networks. If you are following the model correctly, there is no such thing as "direct routing" as you mean it here. > Thats what I wanna call Token Ring, Carousel > networking. :) You can only throw your token at the target on the wall > when you come around to it. I think Token Ring could be a lot better if > you could increase the speed at wich the thing passes the token around. You'll have to ask God to increase the speed of light. > If > each machine is allowed to send or recieve ONE BYTE of data before the > token gets passed, and the passing is very very VERY fast, then it could be > a nice way to do things. Maybe. It's not just one byte of data, it's an entire "frame" (the layer 2 equivilent of a "packet" (roughly)). <> > Ach, the permanence of data thing can easily be dealt with by backing up > valuable data that no one seems to be useing very much. Then just reinsert > the data off of a CD-ROM or something. Hey, I have an idea, you guys > should build a special kind of technique into freenet so that you can > provide an optional ARCHIVE for a freenet node to get stuff from. <> This has been discussed to death. The problem isn't the actual archiving. The problem is that nodes lose the referances to the files and thus have no way of routing to the correct machine that holds the data. It has been conjectured that having permenant data on Freenet while keeping within the project's goals is impossible. - -- Real programmers don't eat quiche. They eat Twinkies and Szechwan food. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.0.4 (GNU/Linux) Comment: For info see http://www.gnupg.org iEYEARECAAYFAjzEpDgACgkQqpueKcacfLQTqwCeI20cL04fRu17cejovpk+IvkL TCUAnA2qkCK/xO8taoWSpgctbZEiqqyr =edTp -----END PGP SIGNATURE----- _______________________________________________ Chat mailing list Chat at freenetproject.org http://lists.freenetproject.org/mailman/listinfo/chat
