On Wed, Jan 22, 2003 at 08:56:12AM -0500, Ed Tomlinson wrote: > On January 22, 2003 08:47 am, Matthew Toseland wrote: > > On Tue, Jan 21, 2003 at 11:17:29PM -0500, Gianni Johansson wrote: > > > On Tuesday 21 January 2003 13:59, you wrote: > > > > > On Tue, Jan 21, 2003 at 06:18:02PM +0100, David 'Bombe' Roden wrote: > > > > > > > > > > On Tuesday, 21. January 2003 18:06, Ian Clarke wrote: > > > > > > Is the "healing blocks" options switched on by default? Obviously > > > > > > it should be or it won't be nearly as effective. > > > > > > > > > > In the unstable build it's not option. It's two fields in the > > > > > download interface. Default (as per gj's code) is 5%, HTL 5. At the > > > > > moment the min-value for the percentage is 0. We could raise it to 5 > > > > > or 10 to enforce healing... > > > > > > > > How is it interfaced? It just runs in the background after the fetch is > > > > finished? > > > > > > The healing blocks are re-inserted after each segment is decoded. In > > > that sense it doesn't run in the background. The data is written before > > > the healing re-insertions, so for single segment splitfiles the user > > > shouldn't notice much impact. single segment == <16Mb > > > > > > For multisegment downloads it does slow things down slightly. > > > > > > > Also, I think the HTL should be raised to 15 or so... > > > > > > My thinking is to leave the default as low as is actually useful, and > > > allow people who want to "sponser" content more actively re-insert at > > > higher htl's. > > > > > > The aggregate effect of people re-inserting even at low htl's should be > > > pretty substantial. Has anyone looked into how healing works in > > > OceanStore? Maybe we could lean on their empirical data to chose > > > re-insertion percentages/htls in a more principled way. 5 was just a > > > guess. > > > > Ewww. That's not how freenet works. Longer HTLs mean the data gets > > closer to the keyspace focus, which means it is MUCH more findable, and > > improves routing. Low HTL reinserts just make it replicate a bit, > > generally where it shouldn't be. Of course I'd defer to Oskar here, but > > that's my assessment. > > If I understand your point correctly, you would rather fewer blocks get > reinserted but that they be inserted with a higher htl so they are placed > where they will be found quickly. Something like, instead healing 10 blocks > at htl 5, do 3 at htl 15...
Actually, the network has been working so well lately that 5 might not be a bad thing... I'm inclined towards 10 though. Compromize: 4% at HTL 7. > > Ed Tomlinson > -- Matthew Toseland toad at amphibian.dyndns.org/amphibian at users.sourceforge.net Full time freenet hacker. http://freenetproject.org/ Freenet Distribution Node (temporary) at http://amphibian.dyndns.org:8889/g9UWlG9KX~w/ ICTHUS. -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: not available URL: <https://emu.freenetproject.org/pipermail/devl/attachments/20030128/c594ae12/attachment.pgp>
