Timestamps are in GMT/BST.
[0:10] * HeIIMind (n=HellMind@) has joined #freenet
[0:21] * HellMind (n=HellMind@) Quit (Read error: 110 (Connection timed out))
[0:21] * HellMind (n=HellMind@) has joined #freenet
[0:26] * NEOatNHNG (n=neo@) Quit (Read error: 110 (Connection timed out))
[0:31] * HellMind (n=HellMind@) Quit (Read error: 60 (Operation timed out))
[0:32] * HellMind (n=HellMind@) has joined #freenet
[0:36] * HeIIMind (n=HellMind@) Quit (Read error: 110 (Connection timed out))
[0:43] * toad_ (n=toad@) Quit (Read error: 104 (Connection reset by peer))
[0:45] * HellMind (n=HellMind@) Quit ()
[1:14] * Krhis (n=Krhis@) Quit (Remote closed the connection)
[1:31] * Malkavian__ (i=Malkavia@) has joined #freenet
[1:32] * dimm0k (n=nmok@) has joined #freenet
[1:33] * Malkavian__ (i=Malkavia@) has left #freenet
[1:33] * dimm0k (n=nmok@) Quit (Client Quit)
[2:11] * Zarggg_ (n=z@) has joined #freenet
[2:18] * gasi_ (n=gasi@) has joined #freenet
[2:24] * Zarggg (n=z@) Quit (Read error: 113 (No route to host))
[2:26] * Zarggg_ (n=z@) Quit ()
[2:34] * gasi (n=gasi@) Quit (Read error: 110 (Connection timed out))
[3:08] * Krhis (n=Krhis@) has joined #freenet
[3:28] * Zarggg (n=z@) has joined #freenet
[3:31] * kalkin-- (n=kalkin-@) has joined #freenet
[3:44] * dimm0k (n=nmok@) has joined #freenet
[3:45] <dimm0k> how do i go about getting some support for requiem?
[3:47] * kalkin- (n=kalkin-@) Quit (Read error: 113 (No route to host))
[4:03] * Zarggg (n=z@) Quit (Read error: 104 (Connection reset by peer))
[4:04] * Zarggg (n=z@) has joined #freenet
[4:08] * Pichu0102 (n=pichu010@) Quit (Read error: 110 (Connection timed out))
[4:09] * Pichu0102 (n=pichu010@) has joined #freenet
[4:13] * Zarggg (n=z@) Quit ()
[4:20] * dimm0k (n=nmok@) Quit ("[BX] Mr. Rogers uses BitchX. Won't you be my neighbor?")
[4:22] * ahuxley (n=ahuxley@) Quit ()
[5:06] * CuriousJohn (n=CuriousJ@) has joined #freenet
[5:19] * CuriousJohn (n=CuriousJ@) Quit ()
[5:25] * kryptos23_ (n=chatzill@) has joined #freenet
[5:42] * kryptos23 (n=chatzill@) Quit (Read error: 110 (Connection timed out))
[5:46] * kalkin-- (n=kalkin-@) Quit (Read error: 104 (Connection reset by peer))
[5:52] * Nande (n=KVIrc@) Quit ("byes, tengan la bondad de ser felices. ;)")
[6:05] * n0ob (n=travis@) Quit ("Now onto something more interesting")
[6:37] * sdiz (n=sdiz@) has joined #freenet
[6:37] * ChanServ sets mode +o sdiz
[7:01] * Hikaru` (n=plop@) has joined #freenet
[7:04] * CrystalDragon (n=marc@) has joined #freenet
[7:04] * CeeDee (n=marc@) Quit (Read error: 54 (Connection reset by peer))
[7:11] * kryptos23_ is now known as kryptos23
[7:13] * Luke771 (n=luke@) has joined #freenet
[7:26] * Pichu0102 (n=pichu010@) Quit (Remote closed the connection)
[7:31] * Pichu0102 (n=pichu010@) has joined #freenet
[7:34] * vi390 (n=fc@) has joined #freenet
[7:43] <FreenetLogBot> r21831 (1155) was built successfully on emu, mirrors are updating
[7:43] <FreenetLogBot> r21832 (1155) was built successfully on emu, mirrors are updating
[7:43] <FreenetLogBot> r21834 (1155) was built successfully on emu, mirrors are updating
[7:55] <FreenetLogBot> The monitoring script has detected a network glitch on http://mirror4.freenetproject.org/ : it has been disabled
[8:10] <FreenetLogBot> The monitoring script has detected that http://mirror4.freenetproject.org/ is eligible to be back on the main mirror rotation list. Welcome Back!
[8:39] * sdiz (n=sdiz@) Quit ("leaving")
[8:43] <FreenetLogBot> r21837 (1155) was built successfully on emu, mirrors are updating
[8:45] * saces (n=saces@) has joined #freenet
[8:45] * ChanServ sets mode +o saces
[8:46] <saces> hi.
[8:47] <saces> i have moved the node to a new box, but it doesnt start: http://code.bulix.org/z6knur-68013
[8:49] <saces> any hints?
[8:52] <nextgens> you've a .bindTo set to your old ip address I bet
[8:53] <saces> yeah.
[8:54] <saces> and the wrapper warning is "normal"?
[8:54] <nextgens> yes
[8:54] <nextgens> hmm no
[8:54] <nextgens> it's not
[8:54] <nextgens> you need the 64 version of the libs
[8:55] * kalkin- (n=kalkin-@) has joined #freenet
[8:55] <nextgens> saces> wget https://checksums.freenetproject.org/latest/wrapper_Linux.zip
[9:05] <saces> ok. is up now. thx.
[9:06] * kalkin- (n=kalkin-@) has left #freenet
[9:14] <saces> how to calculate the max storesize value?
[9:15] <saces> equal to number of 1k-blocks in partition ?
[9:21] * Luke771 (n=luke@) Quit ("[add smart/funny quote here]")
[9:42] <saces> the now box show it very clearly: the external network is the bottleneck ;)
[9:49] <nextgens> saces> I don't get your previous questions
[9:49] <nextgens> what do you mean?
[9:51] * avalonft (n=avalonft@) has joined #freenet
[9:51] <avalonft> hi guys!
[9:52] <saces> nextgens: i have a extra partition for the store dir.
[9:53] <nextgens> you can express it in "size"
[9:53] <nextgens> 1GB is fine
[9:54] <saces> can i set storesize = partition size?
[9:54] <saces> nextgens: its 0.16 TB ;)
[9:56] <nextgens> no, keep some room there too
[9:56] <nextgens> I dunno how much
[9:56] <nextgens> if it runs out of space it can be very bad
[9:57] <nextgens> and you need space for reconstruction anyway
[10:17] * Anarhist (n=volodya@) Quit (Read error: 60 (Operation timed out))
[10:30] * sdiz (n=sdiz@) has joined #freenet
[10:30] * ChanServ sets mode +o sdiz
[10:49] * avalonft (n=avalonft@) Quit ()
[11:01] * infinity0 (n=infinity@) has joined #freenet
[11:01] * cdent (n=cdent@) has joined #freenet
[11:11] * cdent is now known as tr0n
[11:11] * tr0n is now known as cdent
[11:40] * Anarhist (n=volodya@) has joined #freenet
[11:43] * sbc (n=ca@) has joined #freenet
[11:54] * CuriousJohn (n=CuriousJ@) has joined #freenet
[11:57] <FreenetLogBot> The monitoring script has detected a network glitch on http://mirror2.freenetproject.org/ : it has been disabled
[12:12] <FreenetLogBot> The monitoring script has detected that http://mirror2.freenetproject.org/ is eligible to be back on the main mirror rotation list. Welcome Back!
[12:13] * cdent (n=cdent@) Quit (Read error: 110 (Connection timed out))
[12:19] * dbkr (n=nnnnndav@) Quit ("Getting off stoned server - dircproxy 1.2.0")
[12:27] * dbkr (n=nnnnnnda@) has joined #freenet
[12:49] * caytchen (n=caytchen@) has joined #freenet
[13:03] * toad_ (n=toad@) has joined #freenet
[13:03] * ChanServ sets mode +o toad_
[13:08] * nico_32 (n=user@) has joined #freenet
[13:15] * batosai (n=julien@) has joined #freenet
[13:15] * ChanServ sets mode +v batosai
[13:15] <batosai> hi
[13:23] <saces> hi
[13:24] <saces> batosai: can your plugin deal with fms' trust?
[13:24] * greycat (i=rfc1413@) has joined #freenet
[13:25] <nextgens> I need a big key to enqueue
[13:25] <nextgens> any idea?
[13:27] <saces> a linux dvd was postet a few days ago here iirc
[13:44] * Nyquist5 (n=Nyquist5@) Quit ("Leaving")
[13:48] <FreenetLogBot> r21841 (1155) was built successfully on emu, mirrors are updating
[13:54] <batosai> saces: no
[13:54] <batosai> saces: but that shouldn't be too complicated
[13:57] * [Malkav] (i=Malkavia@) has joined #freenet
[13:59] * Myrmica (n=Myrmica@) has joined #freenet
[13:59] <sdiz> anybody using latest trunk?
[14:00] <sdiz> try to play with the configs in /config ...
[14:00] <sdiz> lots of change have been done in the config framework
[14:00] <sdiz> need some tester
[14:09] <nextgens> sanity> I don't like google's code review tool either:
[14:09] <nextgens> it displays the *full* file and not only parts of it: that takes *ages* for my browser on big classes
[14:10] <nextgens> also it adds a footer to the email sent to the lists
[14:10] <nextgens> and html escapes the content of the comments
[14:11] <nextgens> but well, if people want to use it, why not
[14:11] * batosai (n=julien@) Quit ("Ex-Chat")
[14:24] * Myrmica (n=Myrmica@) Quit ("Quitte")
[14:24] * Myrmica (n=Myrmica@) has joined #freenet
[14:24] * Myrmica (n=Myrmica@) Quit (Read error: 104 (Connection reset by peer))
[14:25] * cdent (n=cdent@) has joined #freenet
[14:28] * toad_ (n=toad@) Quit (Read error: 110 (Connection timed out))
[14:35] * toad_ (n=toad@) has joined #freenet
[14:35] * ChanServ sets mode +o toad_
[14:50] <FreenetLogBot> r21842 (1155) was built successfully on emu, mirrors are updating
[14:52] * nextgens has set up a new tool: integration of mantis to svn
[15:01] <toad_> cool, it works now?
[15:01] <nextgens> should
[15:01] <toad_> so is that the same as it used to? picking up the SVN numbers?
[15:01] <nextgens> better than that
[15:01] <nextgens> but I'm still testing it
[15:03] <nextgens> yeah it works :))
[15:03] <nextgens> see https://bugs.freenetproject.org/view.php?id=2499
[15:05] <FreenetLogBot> r21843 (1155) was built successfully on emu, mirrors are updating
[15:06] <toad_> nextgens: nice!
[15:06] <FreenetLogBot> r21844 (1155) was built successfully on emu, mirrors are updating
[15:07] <toad_> nextgens: you should mention it briefly on devl when it's done, just to show ian we're not complete luddites
[15:07] <nextgens> hehe
[15:07] <nextgens> ack
[15:07] <nextgens> do you know how to make "real links" in mantis?
[15:07] <nextgens> possibly with a legend
[15:10] * nextgens is going to keep that format
[15:10] <nextgens> with a link to the review tool to please ian
[15:14] * sdiz (n=sdiz@) Quit ("leaving")
[15:21] <nextgens> toad_> shall I write a detailed explanation or is putting the two regexps in the email enough?
[15:23] <nextgens> well, the regexps are enough I guess
[15:26] <toad_> it doesn't need to be very detailed
[15:28] * n0ob (n=travis@) has joined #freenet
[15:46] <FreenetLogBot> r21846 (1155) was built successfully on emu, mirrors are updating
[15:52] <nextgens> do we want to be notified through FreenetLogBot about status of mantis tickets?
[15:52] <FreenetLogBot> r21847 (1155) was built successfully on emu, mirrors are updating
[15:56] <toad_> nextgens: probably a good ide
[15:56] <toad_> a
[15:57] * Hikaru` (n=plop@) Quit ("Prout")
[16:27] <nextgens> hmmm
[16:27] <nextgens> the regexp doesn't tolerate my bad grammar
[16:28] <FreenetLogBot> r21848 (1155) was built successfully on emu, mirrors are updating
[16:43] <FreenetLogBot> r21849 (1155) was built successfully on emu, mirrors are updating
[16:48] * cdent (n=cdent@) Quit (Read error: 110 (Connection timed out))
[16:52] * Hikaru`_ (n=Plop@) Quit ("Sayonara")
[17:01] <nextgens> sdiz has broken plugin loading
[17:04] * [Malkav] (i=Malkavia@) Quit (Remote closed the connection)
[17:06] * Nande (n=KVIrc@) has joined #freenet
[17:07] * Bombe (n=droden@) Quit (Connection reset by peer)
[17:08] * [Malkav] (i=Malkavia@) has joined #freenet
[17:08] * Bombe (n=droden@) has joined #freenet
[17:08] * ChanServ sets mode +o Bombe
[17:11] <FreenetLogBot> r21850 (1155) was built successfully on emu, mirrors are updating
[17:16] * sdiz (n=sdiz@) has joined #freenet
[17:17] * ChanServ sets mode +o sdiz
[17:23] <FreenetLogBot> r21851 (1155) was built successfully on emu, mirrors are updating
[17:25] * Nande (n=KVIrc@) Quit ("byes, tengan la bondad de ser felices. ;)")
[17:29] <FreenetLogBot> r21853 (1155) was built successfully on emu, mirrors are updating
[17:32] * Nande (n=KVIrc@) has joined #freenet
[17:40] <nextgens> the commit seems to be way faster now that the code-collaborator hook is gone
[17:45] <Tommy[D]> How stable is current trunk? Anything known broken or not working?
[17:45] <nextgens> the config framework is known to be broken
[17:45] <nextgens> but the rest ought to be working
[17:46] <nextgens> and needs testing :)
[17:49] * C1337us (n=St3tson@) has joined #freenet
[17:49] <C1337us> hello
[17:49] <C1337us> Is anyone online?
[17:50] * C1337us (n=St3tson@) Quit (Client Quit)
[17:53] <toad_> rehi
[18:00] <Tommy[D]> huh, he gave us a hole minute to respond, isnt that too much? :)
[18:01] <toad_> hehe he's not used to IRC :)
[18:02] <Tommy[D]> hm, having a very big file in upload with a node crash is bad, now it does delete all the temp data on restart, time consuming
[18:03] <nextgens> Tommy[D]> try trunk
[18:03] <nextgens> it has a feature which will fix it
[18:05] <Tommy[D]> hm, deleting the Anchor file while deleting temp files seems to have no effect, is that intended?
[18:16] <FreenetLogBot> r21858 (1155) was built successfully on emu, mirrors are updating
[18:23] * CuriousJohn (n=CuriousJ@) Quit ()
[18:36] <nextgens> toad_> did you review sdiz's changes yet?
[18:37] <nextgens> plugins seems to be broken and I fail to spot where he introduced the bug
[18:41] * j0sh- (i=null@) has joined #freenet
[18:42] <toad_> nextgens: not yet
[18:42] <toad_> nextgens: i like this idea of a paranoia slider, i have sent a response about it with some ideas
[18:42] <toad_> nextgens: it could replace the opennet question
[18:44] <nextgens> nah
[18:45] <nextgens> they are different things involved here
[18:45] <nextgens> imho we can have different paranoia levels depending on the threat model
[18:45] <nextgens> but those are threat specific
[18:46] <nextgens> for instance, I do care about what my isp can see... but not what a forensic officier can find on my hd
[18:46] <nextgens> because I know that if we reach that point I'm out already
[18:49] * archangelpetro (n=hellequi@) Quit (Remote closed the connection)
[18:49] * archangelpetro (n=hellequi@) has joined #freenet
[18:49] <toad_> nextgens: hmmm maybe
[18:50] <toad_> nextgens: i don't think it's all that multidimensional ... at least not until you get to the deep geek stage
[18:50] * archangelpetro (n=hellequi@) Quit (Remote closed the connection)
[18:50] <j0sh-> lol
[18:50] <toad_> if you know what you're doing, then you should set all the config settings yourself
[18:51] <toad_> but as a rough estimate of where people are at, it makes a lot of sense
[18:51] <toad_> "do i care about datastore seizure" is simply a flag anyway
[18:52] <toad_> if there are only one or two such flags then we can ask about them on the same screen
[18:52] <nextgens> that's the idea
[18:53] <toad_> of course, this can long term be combined with a trust scale for darknet peers
[18:54] <toad_> so if your paranoia level isn't too high, you can do direct (non-tunneled) bloom filter match lookups on trusted peers
[18:55] <toad_> for instance
[18:56] * caytchen_ (n=caytchen@) has joined #freenet
[18:57] * caytchen (n=caytchen@) Quit ("Verlassend")
[18:59] <nextgens> toad_> is there any reason why you don't use the java5 jar of db4o?
[19:00] <nextgens> toad_> I'm willing to release 1156 when the config framework bugs are fixed; any objection?
[19:09] <toad_> of course you have to be careful - just because somebody is trustworthy doesn't necessarily mean their computer hasn't been rooted
[19:13] <nextgens> Bombe> do you have an example where max-size doesn't work?
[19:18] <nextgens> does anyone have a link to a legal file over 2M recently inserted?
[19:18] <FreenetLogBot> r21861 (1155) was built successfully on emu, mirrors are updating
[19:31] <nextgens> Bombe> worksforme
[19:32] <nextgens> I'll commit a cosmetic change but that's about it
[19:36] * sich (n=sich@) has joined #freenet
[19:42] <FreenetLogBot> r21862 (1155) was built successfully on emu, mirrors are updating
[19:45] * makomk (n=aidan@) Quit (Read error: 104 (Connection reset by peer))
[19:53] * HyperDimensions (n=HyperDim@) Quit (Read error: 110 (Connection timed out))
[20:02] * HyperDimensions (n=HyperDim@) has joined #freenet
[20:05] <FreenetLogBot> r21863 (1155) was built successfully on emu, mirrors are updating
[20:05] <FreenetLogBot> r21864 (1155) was built successfully on emu, mirrors are updating
[20:13] * makomk (n=aidan@) has joined #freenet
[20:16] <FreenetLogBot> r21867 (UPnP) was built successfully on emu, mirrors are updating
[20:18] * lattt (n=lattt@) has joined #freenet
[20:25] * NEOatNHNG (n=neo@) has joined #freenet
[20:31] <FreenetLogBot> r21868 (1155) was built successfully on emu, mirrors are updating
[20:38] * CrystalDragon (n=marc@) Quit (Remote closed the connection)
[20:42] * nico_32 (n=user@) Quit ("Ex-Chat")
[20:51] * lattt_ (n=lattt@) has joined #freenet
[20:56] * vi390 (n=fc@) Quit ("This computer has gone to sleep")
[21:01] * lattt_ (n=lattt@) has left #freenet
[21:07] * sich (n=sich@) Quit ("KVIrc 3.2.6 Anomalies http://www.kvirc.net/")
[21:09] * lattt (n=lattt@) Quit (Read error: 110 (Connection timed out))
[21:15] * greycat (i=rfc1413@) Quit ("This time the bullet cold rocked ya / A yellow ribbon instead of a swastika")
[21:23] <toad_> nextgens: moving to the java5 version of db4o is on the todo list
[21:24] <FreenetLogBot> r21874 (1155) was built successfully on emu, mirrors are updating
[21:26] * cdent (n=cdent@) has joined #freenet
[21:27] <nextgens> toad_> I suggest you do it before optimizing the code
[21:28] <nextgens> performances are likely to be very different on it
[21:28] <nextgens> Tommy[D]> trunk is now testable
[21:29] <FreenetLogBot> r21875 (1155) was built successfully on emu, mirrors are updating
[21:30] <nextgens> can testers please upgrade and test this one extensively?
[21:30] * nextgens would like to release it
[21:35] <FreenetLogBot> Emu wasn't able to build r21876 -- see error message on http://emu.freenetproject.org/~svn-build/build-r21876.txt
[21:36] <nextgens> doh; :@ cache
[21:41] * Jase (i=jase@) Quit (Remote closed the connection)
[21:49] * caytchen_ (n=caytchen@) Quit (Read error: 104 (Connection reset by peer))
[21:50] <FreenetLogBot> r21877 (1155) was built successfully on emu, mirrors are updating
[22:00] * cdent (n=cdent@) Quit ()
[22:02] <nextgens> dbkr> are you taking any action on http://archives.freenetproject.org/message/20080809.224354.1fe0fab1.en.html ?
[22:02] * mikusr (i=mikusr@) has joined #freenet
[22:05] <FreenetLogBot> The monitoring script has detected a network glitch on http://mirror1.freenetproject.org/ : it has been disabled
[22:12] <toad_> nextgens: i disagree
[22:13] <toad_> nextgens: the optimisations i've been doing eliminate vast amounts of database work
[22:13] <toad_> nextgens: the 1.5 version is unlikely to be much faster than the 1.2/1.3/1.4 version, but 7.2 may be significantly faster than 6.4; that is something that should be investigated soon
[22:13] * Zarggg (n=z@) has joined #freenet
[22:16] * cdent (n=cdent@) has joined #freenet
[22:21] * FrinkC (n=FrinkC@) Quit ("Bye")
[22:22] <FreenetLogBot> The monitoring script has detected that http://mirror1.freenetproject.org/ is eligible to be back on the main mirror rotation list. Welcome Back!
[22:24] <toad_> http://wiki.freenetproject.org/WarningTextDraft
[22:24] <toad_> any thoughts?
[22:24] <toad_> this will replace the opennet choice page
[22:24] <nextgens> I don't like it
[22:24] <FreenetLogBot> r21879 (1155) was built successfully on emu, mirrors are updating
[22:24] <toad_> have you seen it after my edits?
[22:25] <toad_> what do you not like about it?
[22:26] <nextgens> it's "threat level" focused wheras it should be "threat-model" focused
[22:26] <FreenetLogBot> r21880 (1155) was built successfully on emu, mirrors are updating
[22:26] <toad_> i still don't see the difference
[22:26] * Jase (i=jase@) has joined #freenet
[22:27] <nextgens> let's take an example: resistence to seizure
[22:28] <nextgens> you would disable the bucket encryption & datastore double-encryption only on the last security level (worst)
[22:28] <nextgens> and that would enable opennet
[22:29] <nextgens> it doesn't fit my usecase: I don't want opennet but I don't care about the data which can be gathered through a seizure
[22:29] <nextgens> those are really two different things:
[22:30] <nextgens> "vulnerability to sybil attacks" and "forensic analysis of your HD"
[22:30] <nextgens> they deserve to be handled separately
[22:30] <nextgens> hence we can't only have a "threat level" setting
[22:30] <nextgens> we should identify "kinds" of threats
[22:30] <nextgens> and have one setting for each
[22:31] <nextgens> of course that means we will have to explain to the user each of the individual threat type
[22:31] <nextgens> but well, it's worth it
[22:31] <toad_> well imho 99% of our threat model is remote network attackers
[22:31] <nextgens> knowledge is always good to have
[22:32] <nextgens> and that's also a marketing argument: once you've explained to the user what threats freenet protects him from, you can help him comparing it to "other anonymity products"
[22:32] <nextgens> toad_> then why do we even care about seizure?
[22:32] <nextgens> it *does* slow things down
[22:32] <toad_> nextgens: we can't explain every paranoia option to every new user
[22:33] <toad_> it's not remotely realistic to do so
[22:33] <toad_> they wouldn't get half way through any such explanation
[22:33] <nextgens> I'm not talking about individual options here
[22:33] <nextgens> I'm talking about "kind of vulnerabilities and threats"
[22:33] <toad_> well, what are you talking about? afaics there are only two dimensions here: seizure and network
[22:33] <nextgens> which is very different
[22:34] <nextgens> those are the two obvious ones
[22:34] <nextgens> but I'm sure we can find more if we brainstorm for 5mins
[22:34] <toad_> seizure is a simple boolean: "Click here if you don't care about what traces Freenet leaves on disk"
[22:34] <nextgens> my brain refuses to do so past midnight but maybe yours is willing to :p
[22:34] <toad_> vulnerability to your Friends is a big one, but only applicable if you have Friends in the first place, and IMHO it will vary by peer
[22:34] <nextgens> toad_> yes exactly hence we can ask that question
[22:35] <toad_> nextgens: we can add that one on, sure, but we can't add 10 of them on
[22:35] <nextgens> agreed but as you said, they are hardly two :)
[22:35] <toad_> imho that one belongs in an Advanced Options ... box
[22:36] <nextgens> I don't think so
[22:36] <nextgens> freenet has to be secure out of the box
[22:36] <nextgens> or at least to propose to be secure
[22:36] <toad_> it would default to "secure", of course
[22:36] <toad_> unless the user chooses not-secure-at-all
[22:36] <nextgens> and as such an option has a huge performance impact, enabling it by default (what we have been doing 'till now) is not an option
[22:36] <toad_> it does?
[22:37] <toad_> if so we should look into options for doing something about it
[22:37] <toad_> i thought you had decided it wasn't a big deal
[22:37] <nextgens> especially if you plan to deploy nodes on gamer's playstations!
[22:37] <toad_> you implemented the first-few-buckets-in-RAM thing, right? how much performance gain do you get from a 10MB pool?
[22:37] <nextgens> I haven't finished investigating the bucket thingy
[22:37] <nextgens> but the gains are already there
[22:38] <nextgens> we don't need 10+mins to clear brazillions of left over tempbuckets anymore
[22:38] <toad_> imho we can probably afford a 10MB pool
[22:38] <nextgens> that's on by default on trunk
[22:38] <toad_> nextgens: i don't see what clearing temp buckets has to do with encryption
[22:38] <nextgens> now the question and parameter to tweak is the max-size admissible in the pool
[22:38] <nextgens> (per object)
[22:38] <nextgens> right now it's Short.MAX_VALUE
[22:38] <toad_> nextgens: have you implemented aging?
[22:39] <nextgens> no, but I should have
[22:39] <toad_> imho we should dump stuff after some period, maybe a few minutes
[22:39] <toad_> that is, dump it to disk
[22:39] <nextgens> that's next step: we will need stats on the pool
[22:39] <toad_> also imho Short.MAX_VALUE is too small
[22:39] <toad_> many web pages and most images will be over that
[22:39] <nextgens> toad_> it has to do with performances
[22:40] <nextgens> toad_> and paranoia too to some extend
[22:40] <FreenetLogBot> The monitoring script has detected a network glitch on http://mirror2.freenetproject.org/ : it has been disabled
[22:40] <toad_> nextgens: well then only dump it if the pool is full
[22:40] <toad_> but it's probably better to dump it in advance
[22:40] <toad_> anyway, if this is giving us a massive performance gain, something is seriously wrong with our implementation
[22:40] <toad_> think about it: how many layers of decryption do we do before we even reach temp buckets?!
[22:41] <nextgens> the bottleneck here is IO
[22:41] <toad_> i suppose if we fetch stuff from the archive cache, not many - but even then we have to fetch and decrypt enough to reach the point of accessing it
[22:41] <nextgens> before we were using file buckets for everything
[22:41] <nextgens> we aren't anymore
[22:41] <nextgens> that's the big perf. gain
[22:41] <toad_> nextgens: you said it was crypto
[22:42] <nextgens> well, I've implemented both of them in a row
[22:42] <toad_> nextgens: you said it was MUCH faster if we just turn off crypto and *keep them on disk anyway*
[22:42] <nextgens> so I didn't test them individually
[22:42] <toad_> ah, ok
[22:42] <nextgens> yes, now I recall it
[22:42] <toad_> well then you need to find out which is more important
[22:42] <nextgens> I said that fproxy and the content filter were faster
[22:42] <toad_> a good way to improve the speed of the content filter is to filter stuff on insert
[22:43] <nextgens> we can't do that for security reasons
[22:43] <toad_> why not?
[22:43] <nextgens> we need a filter on the recipient's side too
[22:43] <toad_> we'd run the content filter anyway
[22:43] <toad_> but if it's already filtered, we won't need to copy it
[22:43] <toad_> we only need to read it
[22:43] <nextgens> well, is it way faster to run it when it doesn't throw?
[22:43] <nextgens> I'm not sure
[22:44] <toad_> if it throws we have to run it twice, and allocate a bucket, copy the data
[22:44] <nextgens> okay
[22:44] <nextgens> well, I know what I've to do then :)
[22:44] <toad_> :)
[22:44] <toad_> please do find out whether it's crypto or I/O that is the problem also
[22:45] <nextgens> we need wider testing
[22:45] <nextgens> hence I was willing to release a new build
[22:45] <toad_> I wouldn't expect crypto to be a big deal on modern systems
[22:45] <nextgens> with the pool and the crypto enabled
[22:45] <nextgens> so we can get selected testers to disable crypto and compare performances
[22:45] <toad_> so buckets still encrypted, but a limited RAM pool? that's fine by me
[22:46] <toad_> well we can get selected testers to test trunk
[22:46] <nextgens> that's current trunk's configuration
[22:46] <toad_> yeah except it sucks cos the limits are wrong
[22:46] <nextgens> that's what I've been trying to do over the past two days :)
[22:46] <toad_> very little useful content will be under 32K
[22:46] <nextgens> well
[22:46] <toad_> if you have 10MB of pool to work with, imho a limit of 128K is reasonable
[22:47] <nextgens> it seems enough to speed fproxy up
[22:47] <nextgens> well, increase it then :)
[22:47] <toad_> exactly what performance are we talking about here?
[22:47] <toad_> reloading a page you've already loaded?
[22:48] * nextgens does it
[22:48] <toad_> you definitely need migration anyway, or the archive extraction cache will fill up RAM and then everything will have to go to disk
[22:48] <nextgens> they are lots of stuff which are 32K
[22:48] <toad_> those buckets are long-lived
[22:49] <nextgens> fec blocks are, aren't they?
[22:49] <toad_> fec blocks are exactly 32K indeed
[22:49] <toad_> but Short.MAX_VALUE is 32K-1
[22:49] <nextgens> that's where the bigGain is I think
[22:49] <toad_> hmmm, that's possible
[22:49] <nextgens> it's not Short.MAX_VALUE then :)
[22:49] <nextgens> it's 32KiB
[22:49] <toad_> ok
[22:50] <toad_> well, either you need timed/thresholded migration, or you need a flag for long-lived temp buckets
[22:50] <toad_> imho the former is the better option
[22:51] <toad_> you could even do it like paging - migrate stuff after 2 minutes but keep it in RAM, marked clean, until it gets dumped to make way for other stuff
[22:51] <FreenetLogBot> r21881 (1155) was built successfully on emu, mirrors are updating
[22:51] <toad_> and have LRU of course
[22:52] <nextgens> we don't want to have too much overhead either
[22:52] * toad_ is surprised the OS caching is so ineffective though ... unless it's all the layers of crypto, like you said ...
[22:52] <nextgens> or the brazillion of threads we use wich mess the prediction up
[22:52] * toad_ supposes caching in userspace is generally a good idea ... apache gained massively from it for example
[22:53] <toad_> well that's something we should deal with too
[22:53] <nextgens> I thought your branch did address that too
[22:53] <toad_> what? the number of threads? no
[22:54] <nextgens> isn't the db-thread serialized?
[22:54] <toad_> yes
[22:54] <toad_> so is datastore access
[22:54] <toad_> but client layer stuff was mostly serialised before anyway
[22:54] <nextgens> and you kept the forking at the client-layer's level?
[22:54] <nextgens> where do we fork then?
[22:54] * nextgens is lost
[22:55] <nextgens> it's not on fec decoding either
[22:55] * toad_ pastes on #freenet-chat
[22:56] <toad_> some of it is single thread jobs which probably should be moved to the ticker; some of it is data transfer; a lot of it is just requests
[22:56] <FreenetLogBot> The monitoring script has detected that http://mirror2.freenetproject.org/ is eligible to be back on the main mirror rotation list. Welcome Back!
[22:56] <nextgens> and can't we serialize requests?
[22:56] <toad_> if FOAF halves average request round trip time, will that mean more, less, or the same number of parallel requests?
[22:56] * nextgens was working on that at some point
[22:57] <nextgens> more
[22:57] <toad_> serialize RequestSender ? not a good idea imho, end up with complex hard to debug state machines
[22:57] <nextgens> because they will finish earlier... and users/nodes will handle more
[22:57] <nextgens> the network is condamned to handle more and more requests
[22:58] <toad_> why is that bad?
[22:58] <nextgens> it's not but we can't base our design on the fact that there will be less requests because they are more efficient:)
[22:58] <toad_> :)
[22:58] <toad_> ok
[22:59] <toad_> well... we could shift to state machines, but you'd need to convince both me and ian that the state machine is as simple and easy to debug as the original threaded implementation
[22:59] <toad_> or we could import some continuations library from RIFE or somewhere
[23:00] <toad_> one of the big things behind the 0.7 rewrite was a drive to simplify the low level request logic by using simple blocking threaded requests
[23:01] <nextgens> state machines are trivial to debug when there is no jvm bug
[23:02] <nextgens> because interractions in between them can be modelised
[23:02] <nextgens> you can proove a state-machine design
[23:02] <toad_> i haven't seen much formal analysis going on here recently
[23:02] <nextgens> (they are even abstraction languages to describe them)
[23:03] <toad_> it needs to be practically easy, not just theoretically neat
[23:03] <nextgens> toad_> that's something I've learnt to do at university :)
[23:03] <nextgens> often used to deal with real-time things
[23:04] <nextgens> wich is kind of something we do need to
[23:04] <nextgens> and we don't have right now
[23:04] <nextgens> some requests timeout and we have no idea why
[23:04] <toad_> you think?
[23:04] <nextgens> some messages are unclaimed and we don't know where they got lost
[23:05] <nextgens> I'm not convinced that blocking threads are any easier to debug
[23:05] <toad_> our requests are pretty simple
[23:05] <nextgens> but that's an other story :)
[23:05] <nextgens> sure, but they are buggy
[23:05] * nextgens has seen timeouts even on simulations
[23:06] <nextgens> anyway
[23:06] <nextgens> back to the original topic:
[23:06] <toad_> timeouts in simulations are caused by cpu overload
[23:06] <nextgens> I don't like your new opennet-warning-design :p
[23:07] <toad_> :|
[23:07] <toad_> why?
[23:08] <nextgens> because from versions to versions we move towards saying that opennet is an acceptable tradeoff
[23:08] <nextgens> whereas it's not
[23:08] <nextgens> it's just like saying that it can be turned off
[23:08] <toad_> well, whether or not to have opennet is a separate question, isn't it?
[23:08] <nextgens> of course it can... but it will be too late
[23:08] <toad_> it depends on your attacker
[23:08] <nextgens> sure
[23:08] <toad_> if they keep packet traces indefinitely, then sure, it'll be too late
[23:09] <toad_> if they harvest *before* legislation is introduced to make freenet illegal, then sure, you're fucked
[23:09] <toad_> but why would they do that?
[23:09] <nextgens> okay: third kind of threat: "shall we worry about mid-term behaviour analysis"?
[23:09] <toad_> of course the intelligence services will, but the cops, the RIAA?
[23:09] <nextgens> in France they did
[23:09] <nextgens> even though it was illegal
[23:10] <toad_> (not that we encourage music piracy!)
[23:10] <toad_> s/RIAA/corporates
[23:10] <toad_> they did what exactly?
[23:10] <toad_> mid-term behaviour analysis?
[23:11] <nextgens> harvest data, sent illegal dmca takedown notices (even though they are not valid here), ...
[23:11] <toad_> ah
[23:11] <toad_> well yeah but that's not the same thing
[23:11] <nextgens> and when the right legal tools have been voted they switched to the "legal way of doing it"
[23:11] <toad_> they were harvesting the users manually sharing specific illegal files
[23:11] <nextgens> asking for commissions rogatoires
[23:13] <toad_> nextgens: as far as i can see there is precisely no chance of building a true darknet organically without growing an opennet first to seed from
[23:13] <nextgens> toad_> ISPs could be keeping stats of who is connecting to FPI's server
[23:13] <toad_> if you have any bright ideas as to how to bootstrap a darknet then let me know
[23:13] <nextgens> viral marketing is the only way to go
[23:13] <nextgens> that's how all social networks have grown
[23:13] <nextgens> but we're far from being ready for that
[23:14] <toad_> how do we make that happen? we gave it over a year to grow organically and all that happened was random reference exchange!
[23:14] <toad_> imho viral growth won't happen until the network is fairly large already, as well as faster
[23:14] <nextgens> and imho we would spend to much time dealing with the stupid (^wcomputer illiterate) users it would bring
[23:14] <toad_> heh
[23:15] <toad_> well, computer illiterate users eventually become computer literate users, and help out those who are still newbies
[23:15] <nextgens> no we didn't give it over a year
[23:15] <toad_> yes we did
[23:15] <toad_> we postponed opennet for a *long* time
[23:15] <toad_> it didn't get us growth, it certainly didn't get us true-darknet growth
[23:15] <nextgens> sure but we didn't address the core problems in the meantime
[23:15] <nextgens> opennet doesn't either
[23:16] <nextgens> it's been around for long now
[23:16] <toad_> one of the core problems was the lack of opennet. the network was dominated by #freenet-refs connections and therefore had a random topology, so it was slow
[23:16] <nextgens> we had a major marketing plan when it went out (something we didn't have during the year of darknet growth)
[23:16] <toad_> when we introduced opennet, the network sped up dramatically, at the same time as significantly growing
[23:16] <nextgens> and no, the network isn't growing any faster now
[23:16] <nextgens> toad_> you're mixing things up here
[23:17] <nextgens> we did fix the implmentation in the meantime
[23:17] <nextgens> we are comparing oranges and eggs here
[23:17] <nextgens> how many critical bugs did we fix since then?
[23:17] <toad_> sure, but the biggest thing that changed over the few months around opennet's introduction, in terms of core stuff, was opennet
[23:17] <toad_> and we had a major performance gain over that period
[23:17] <toad_> and we know precisely why: 99% of the network was random, so the topology sucked
[23:17] <nextgens> no
[23:17] <toad_> random because of #freenet-refs
[23:18] <toad_> which is something we could not prevent without opennet
[23:18] <nextgens> we use various constants grown from alchemy... and they are tied to network size
[23:18] <toad_> hmmm?
[23:18] <nextgens> imho the network wasn't big enough to work properly, that's all
[23:19] <nextgens> we should have changed a few constants and that would have helped
[23:19] <toad_> i doubt that very much
[23:19] <nextgens> we are talking about a factor of 10 in terms of network size
[23:19] <toad_> growing normally results in *LESS* performance
[23:19] <nextgens> in the meantime we didn't change those constants
[23:19] <toad_> you have scaling costs
[23:19] <toad_> except for very small networks where we get problems with RNFs etc
[23:20] <nextgens> unless you use silly set and innapropriate constants
[23:20] <nextgens> like our hardcoded htl value
[23:20] <toad_> well yeah, and a factor of 10 growth is another very good reason to have opennet
[23:20] <nextgens> it can't possibly work on a small network
[23:20] <toad_> an anonymous network with only 100 users is not anonymous
[23:20] <nextgens> you've filled in a ticket about that not long ago, remember?
[23:20] <toad_> an htl of 10 won't work well on a network with less than 10 nodes, sure
[23:20] <toad_> you get RNFs
[23:21] <nextgens> it's not much more with 4000nodes
[23:21] <toad_> it's 40 times more, no? :)
[23:21] <nextgens> 4k nodes can be simulated locally
[23:21] <nextgens> needless to say that anyone can do a sybil attack on a such a network
[23:21] <toad_> sure
[23:22] <toad_> but try doing that on a 400k node network
[23:22] <nextgens> especially now that opennet has been deployed
[23:22] <toad_> even on opennet
[23:22] <nextgens> on the 400nodes network it wasn't possible
[23:22] <toad_> imho 400k node network is achievable if we improve performance and so on
[23:22] <toad_> not true
[23:22] <nextgens> by the time the network has grown to 400k nodes ISPs will be able to do sybil attacks on it
[23:22] <toad_> most users on the 400 node network had to come back every month or so for more noderefs
[23:23] <toad_> for a long period until they got "good" refs
[23:23] <toad_> the maintenance cost was unbelievable
[23:23] <nextgens> are you talking about those silly users who don't connect to people they know?
[23:23] <nextgens> we got it wrong marketing wise
[23:23] <toad_> yes, making up 90-99% of the total users
[23:23] <nextgens> we said "you need 20 refs"
[23:23] <nextgens> which wasn't true
[23:24] <toad_> it was true for good performance
[23:24] <nextgens> we needed that many because the implementation wasn't up to the level
[23:24] <toad_> also when you factor in the fact that most nodes are not online 24x7
[23:24] <nextgens> that's a permanent problem
[23:24] <toad_> exactly, it's not something we can solve, it's something we have to work around
[23:24] <nextgens> I'm curious and would like to know what's the average uptime on opennet
[23:25] <nextgens> I bet that the churn is so bad that we probably won't ever be able to do anything with it
[23:25] <toad_> on my node, 0, 38, 51, 74, 0, 63, 0, 8, 99
[23:25] <toad_> i dunno what the zeros are about
[23:25] <nextgens> right now the network works well because there is very little data and everyone is requesting the same
[23:25] <nextgens> so caching mechanisms are performing well
[23:25] <toad_> not true
[23:25] <nextgens> ?
[23:26] <toad_> your stats have shown that we have specialisation, and that we serve in our specialisation and not out of it
[23:26] <nextgens> toad_> what's your current cache hit? :)
[23:26] <nextgens> here it's above 30%
[23:26] <nextgens> and I've a small store/cache
[23:26] <toad_> 3% :(
[23:26] * cdent (n=cdent@) Quit (Read error: 110 (Connection timed out))
[23:26] <nextgens> then we need to fix the stats
[23:26] <toad_> mind you i have a lot of local requests
[23:27] <nextgens> and we shouldn't cache data within our specialisation
[23:27] <toad_> remote requests are 5%, local requests are 8%
[23:27] <nextgens> how big is your store?
[23:27] <toad_> nextgens: we cache whatever goes through us. whatever goes through us is generally stuff close to our location.
[23:27] * nextgens smells the 32MB ds
[23:28] <toad_> 1.59GB cache, but as i said lots of local requests ... deleted it before starting up and starting lots of local requests
[23:28] <toad_> so it's probably mostly stuff that my node requested
[23:28] <nextgens> toad_> it's silly to have the data both in the cache and in the datastore provided that the cost of access and replacement policy are the same
[23:28] <toad_> nextgens: i don't see why a block would be in both the store and the cache
[23:28] <toad_> anyway my store is only 210MB
[23:29] <nextgens> we get an insert request in our specialization: we store it in the store
[23:29] <nextgens> a request: we get a cache fail, we serve the data and push into the cache
[23:29] <toad_> eh?
[23:29] <nextgens> we serve from the cache...
[23:29] <toad_> we check the store first
[23:29] <nextgens> lru gets rid of the data in the store when it's full
[23:29] <toad_> if it's in the store, we return it, end of story
[23:29] <nextgens> and we keep on serving from the cache :/
[23:30] <toad_> then we check the cache, if it's in the cache we return it
[23:30] <toad_> if we didn't serve first from the store, that would indeed be a bug
[23:30] <nextgens> okay
[23:30] <nextgens> usually you do serve from cache because the cost of access is smaller than on the store
[23:30] <nextgens> hence I assumed that's what we did
[23:31] <toad_> usually we serve from cache because the data is more likely to be in the cache at the moment
[23:33] <toad_> anyway ... given that we won't get rid of opennet ... any comments on http://wiki.freenetproject.org/WarningTextDraft ??
[23:34] <nextgens> why is it more likely to be there?
[23:34] <toad_> nextgens: a very good question
[23:34] <nextgens> because there isn't much data on the network
[23:35] <toad_> the only answer i have at the moment is that it fills up very slowly
[23:35] <toad_> and therefore doesn't get many hits, because we change location
[23:35] <nextgens> it shouldn't if requests were sharply specialized
[23:35] <toad_> nextgens: it doesn't cache requests
[23:35] <nextgens> ah, here we go: location churn
[23:35] <toad_> nextgens: it only caches inserts, and then only if we are closer than any of our peers
[23:36] <toad_> right, location churn. imho location churn is caused by pure opennet nodes engaging in swapping. they shouldn't.
[23:36] <toad_> hopefully vive will come up with an answer to this
[23:36] <nextgens> any news from the simulation front?
[23:37] <toad_> well he hasn't posted anything on the topic since his re-hi mail relatively recently
[23:37] * sbc (n=ca@) Quit ("Ex-Chat")
[23:37] <nextgens> ok
[23:38] <nextgens> toad_> well, as I told you, I articulate the warning around two kinds of threats
[23:38] <nextgens> toad_> well, as I told you, I would articulate the warning around two kinds of threats
[23:38] <toad_> which two?
[23:38] <nextgens> network-based threats and local-threats
[23:38] <nextgens> because the user can easily understand both of them
[23:39] <nextgens> maybe we need a third one which would be trachery
[23:39] <toad_> so peers doing correlation attacks ... yeah, that's difficult to deal with ...
[23:39] <nextgens> that's tunnels
[23:40] <nextgens> and shared bloom-filters on the other side of the slider :)
[23:40] <nextgens> but right now we could put FOAF in the balance
[23:40] <toad_> hmmm maybe
[23:40] <nextgens> so yeah, I would say 3 kind of threats
[23:41] <toad_> i was under the impression that we wanted to enable FOAF even on opennet, where treachery is assumed
[23:41] <nextgens> those are easily understandable by any user
[23:41] <nextgens> me too
[23:41] <nextgens> but well, we are foolish, aren't we?
[23:41] <nextgens> publishing our peer's locations is probably something we want to do on opennet
[23:42] <toad_> well, we do have some defence against it even on opennet
[23:42] <nextgens> the jokes we call "mitigation techniques" I am supposed to have implemented?
[23:43] <toad_> yeah the limit to 20 peers thing
[23:43] <nextgens> on opennet you just get more links
[23:43] <nextgens> which is trivial
[23:43] <nextgens> sybil Rules
[23:43] <toad_> okay, so it IS a treachery issue, but we enable it on opennet anyway as there are easier attacks if you're targeting a single peer
[23:43] <toad_> err a single node
[23:44] <toad_> .... but on darknet, it should be linked to the peer trust level?
[23:44] <nextgens> to the general "treachery sensivity level"
[23:45] <nextgens> hmm
[23:45] <nextgens> we need to agree on some terminology here
[23:45] <nextgens> what do you call "peer trust level"?
[23:45] <nextgens> is that what I've just called "treachery sensivity level" or a per-peer setting?
[23:56] <toad_> welwell, to both
[23:56] <toad_> gaaaaaaah
[23:56] <toad_> sorry, system stability issues
[23:57] <toad_> imho the 3 axis suggestion makes sense
[23:57] <nextgens> sensivity
[23:57] <nextgens> is that english?
[23:57] <toad_> but the treachery sensitivity level imho should probably be strictly per peer
[23:57] <nextgens> I'm replying to ian
[23:57] <nextgens> sensitivity
[23:57] <nextgens> okay
Irc logs of #freenet : 2008 2007 2006 2005
These logs were automatically created by FreenetLogBot on chat.freenode.net using the Java IRC LogBot.