<snip>
However, with what we have learned over the past few months, we need to consider that NGR isn't ready for primetime and just throwing it in at
the deep-end to see if it would swim hasn't really worked.
There is also a cold financial reality here - we are barely keeping afloat with my regular appeals for donations to our existing userbase, and it is only going to get more difficult as time goes by. Only a release can generate sufficient donations to get us away from this precipice where we almost run out of funding every few weeks.
Thus we need a release, we have a lot of innovation in the current codebase that isn't tied to NGR (OCO, NIO etc), but the release can't contain NGR until it is working properly. I think, therefore, it is logical to spend a bit of time ensuring that the old routing still works, and produce a release on its own network that is the default people download. It makes sense for our users, it makes financial sense, and it relieves the core developers of the time-pressure to make NGR work - this is a pressure that I certainly haven't enjoyed and I suspect that the same is true of the other devs.
Everything here is up for discussion, if the consensus is that we should give NGR more time to sort itself out then I can live with that, but we need to take a step back and have this meta-discussion or risk lending weight to those who accuse us of ignoring our users.
Here's where I think we stand w.r.t. getting NGR working:
1. We probably should go with Toad's idea of dropping the HTL system in favor of a timeout-based system. This is a radical change, and needs to be thought-out a bit more before it is implemented. However, I would think we should be able to finish the design "on paper" in less than a week.
2. We should implement ian's back-off scheme, Unobtanium routing, and a corrected estimate() formula.
3. Once all that is done, we will see whether or not the problem of too-many-transmitting-connections has been solved. (My own personal feeling is that we'll still have that problem and that we should go with something like QR-based-on-bw or QR-based-on-send-queue-length... but we shall see.)
So, I think we are on the verge of NGR working well.
But the bottom line question: how long will it take to do the above? Take these guesses with a grain of salt: 1 week for (1), 1 week for (2), 1 week for (3). If those guesses really hold up and I'm right that NGR will work after that, then it probably makes sense to "give NGR more time to sort itself out". But, I can assure you from past experience, my guesses are probably wrong and you should double them!
Given the above and the cold financial reality you mentioned, I would not be opposed to another release w/o NGR, if it would bring in more cash and give Toad a bit more time to implement the solutions. He's fast, but we don't want him put under pressure.
-Martin
_______________________________________________ Devl mailing list [EMAIL PROTECTED] http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl
