* Matthew Toseland <toad at amphibian.dyndns.org> [2008-02-29 11:43:03]:
> Ian has stated that Frost is a separate project, and we should not wait for a > usable FMS. (Background: Frost has been severely DoS'ed recently due to being > based on KSK queues, and FMS is the answer). Right now we are not actually > being DoS'ed, but yesterday we were; the spammer occasionally takes a day > off, but IMHO he's likely to stop spamming now until shortly before we ship > 0.7.0, and then make Frost unusable again. > I don't think that fms is *the* answer; it's just part of what could be one answer. I'm still sceptical on how introduction works and I haven't seen any good description of the trust-ranking algorithm yet. > 1) We could just not ship Frost, in order to meet Ian's totally arbitrary > deadline (not that totally arbitrary deadlines aren't of some value!). > However IMHO we will have *dramatically* less user retention if we don't have > a usable chat client - at least on the order of 50%. This is because Freenet > is a community, and no community can function without chat. > Even though I see some perfectly valid reasons to drop frost (namely the fact that it uses nameless keys and that it is slow) I don't think that dropping it because it's spammed is a valid one. That would probably piss frost developers off (something no one has any interest in doing) and wouldn't help building a community either ... Not an option as we don't have any alternative yet. The community will have to bear with spam for the time being. > 2) Or we could ship Frost even though we know it will probably be DoS'ed > again, and may be being actively DoS'ed at the point that we ship it. In > which case IMHO we should advise users about this fact in our announcement > for 0.7.0. Maybe some good publicity will come from it - it could hardly be > any worse than us shipping an application which is utterly useless, and each > user having to figure that out themselves! Telling users will put them > off ... but not telling users will also put them off. > I'm in favor of that one: informing users hardly ever hurts. > 3) Or we could make a working chat system of some kind a release blocker, and > act accordingly: review third party code and help in porting FMS to java if > that is necessary. At the moment FMS is written in C (and therefore not > bundle-able), and uses a combination of HTTP and NNTP interfaces (and > therefore is not user friendly). IMHO the critical path would be to translate > it to java, implement it as a plugin, implement a separate plugin with a web > interface based on that of Worst, bundle those, and let the Frost devs port > Frost to use the plugin. Some of this has already been started, but I haven't > seen much progress recently on the FMS board (which oddly is the only board > never to have been DoS'ed). > I agree with Ian: it shouldn't be a release blocker. We don't even have a good design for an alternative : needless to say the spammer will be long bored before we get a working implementation of it ;) You forgot one option here: What about bundling syndie ? Its design is spam-proof and it has the moderation features some vocal people from the community have been asking us for a long time. NextGen$ PS: Frost is the reference implementation of the current messenging system but others are already available (Thaw with MiniFrost, Worst,...). I think you should mention them too from time to time (their authors deserve some credit as well). -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 189 bytes Desc: Digital signature URL: <https://emu.freenetproject.org/pipermail/devl/attachments/20080229/0c309a25/attachment.pgp>
