On 03/12/15 00:27, Psalle wrote: > On 03/12/15 00:44, Matthew Toseland wrote: >> On 02/12/15 22:34, Psalle wrote: >>> Let me hijack the topic at this point (following Victor Denisov trail) >>> into another though experiment: instead of arguing what Freenet needs >>> or could do, I'll contribute my *chief* reason not to use it (I've >>> been running it from time to time since around 2005). >>> >>> Although I just said I'll say one chief reason, I'll cheat and offer >>> one using three different hats. >>> >>> As a /user/, freenet is extraordinarily heavy for the computers I've >>> used it on (some of them not that low spec). Disk trashing in some of >>> them was so interfering with normal use to make it unbearable. >> Unfortunately we've done nearly everything we can about this. The client >> layer is *far* more efficient than it was on disk I/O, and the datastore >> too. >> >> Maybe it needs some careful profiling. >> >> One thing we could do is use I/O priorities though, especially on >> Windows. This might help significantly. > Curiously, my subjective impression is that I/O load is more > noticeable on linux than on windows (not for freenet but in general). > I have had problems with very large dropbox folders in recent years > but only when using ubuntu.
Well, Linux also has system calls for I/O priorities. There is a library for calling syscalls which we could bundle and use, and there is a pull request for using this. >>> As a /programmer/, the two times I've tried to catch a bug or >>> contribute some code, I found it exceedingly difficult to dive into >>> it. Arguably I could have chosen too difficult points of entry... >> Probably. :| >> You should ask for some pointers? > Yep. I don't like nagging people but I guess there's a thing as being > too shy. >>> As a /donor/ circa 2007 IIRC, I felt my monthly contribution was being >>> wasted on features that led nowhere. >> Any particular features? Ancient history I guess... > Well, at that time my sensation was that there was a piling of > tentative ideas on top of ideas in the hope that some of them would > cause a breakthrough, but I felt that even a good idea would get > drowned in the general noise. Mind you, this was a non-programmer > perception, so things were probably different from the inside. Not that different. > > Put another way, I felt that there was not a clear understanding of > why/how well the network as a whole was working, so all the coding > without a solid analysis discouraged me. I loved the stats and > simulations, not surprisingly, so in that regard I was happy. Yep. We try to be a bit more empirical. My project this year should help with this. > > I understand this is a very hard problem from a scientific point of > view. I guess I miss more developments in the theoretical side of > things. It's not that I am against experimental algorithms, but that I > see the freenet codebase too large and complex to really be an > efficient platform for experimenting right now. In that regard, I > admire how much effort you and others have devoted to such a > frustrating problem. Yes, exactly, it is hard to evaluate changes so we can't just try them and see what happens. We need to be able to simulate changes "in vitro" first, and then, if possible, have micro-benchmarks so that we can evaluate specific consequences of our changes on the real network.
signature.asc
Description: OpenPGP digital signature
_______________________________________________ Devl mailing list [email protected] https://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl
