On 02/11/15 13:02, Bob Ham wrote: > On Mon, 2015-11-02 at 06:28 -0500, Steve Dougherty wrote: >> On 11/02/2015 06:26 AM, Steve Dougherty wrote: >>> On 11/02/2015 06:21 AM, Bob Ham wrote: >>>> On Mon, 2015-11-02 at 05:54 -0500, Steve Dougherty wrote: >>>>> one >>>>> may observe that we have not written up protocol documentation, making >>>>> that our current strategy. >>>> I'm not sure what you mean; what are you referring to with the word >>>> "that"? What is your current strategy? >>> The current strategy is to not document things. I think calling it a >>> strategy is a stretch because a strategy requires coordination and >>> planning. No one has wanted to document the protocol, so it hasn't >>> gotten documented. >> Er, that's poor phrasing. No one has wanted to document the protocol >> enough to document it themselves. I'm not being insightful here - it >> hasn't happened, so our "strategy" is to not do it. > You're right, that's not a strategy. > > Regardless, I think there's some confusion here. I'm not talking about > a strategy for creating a file containing information about the Freenet > protocol. What I'm talking about a strategy for getting the protocol to > a point where it's worth other people implementing it and then writing > high quality documentation that allows them to do so, possibly even > publishing an RFC describing it. > > It seems that nobody here believes that in five years' time Freenet > developers will be publishing an RFC. Instead, people seem to be > focussed on updating the website, worrying about user interface niggles > and fretting over whether users can install new versions of Fred easily > enough. Documentation is good.
But multiple implementations just don't make sense until the protocol is somewhere close to "final". Even then they're problematic. Why do you think TCP still relies on packet loss to signal congestion, rather than ECN? Because there are thousands of buggy TCP firewalls which drop packets with the ECN bit set, or worse, home routers which crash completely when they see it! HTTP extensions work because extensions to one server do not affect anyone else. Freenet isn't like that. It's a distributed protocol, more like BGP or Bitcoin. We gain considerably from being able to change stuff without worrying too much about legacy incompatibilities, and given our limited resources, this is a good thing. Furthermore, alternative implementations would not be noticeably faster. They might have slightly lower overheads in memory and CPU, but at this point that's probably not the main problem anyway: If you have a datastore you need to have a certain amount of disk I/O, for example. The bottlenecks are elsewhere, on the whole. > I'm talking about a strategy for getting the project to a point where it > can actually have meaningful impact on the world. There doesn't seem to > be any direction in the project. There's some software and some kind of > community around it but there seems to be no vision of how to move > forward to a point where the project contributes to something wider. Making a difference is not a matter of standardisation. Which had the greater impact, POSIX or Linux? > What is the priority of the project, is it to ensure that as many > Windows users as possible have a little Freenet icon in their status > tray? Or is it to play a role in creating a world where nobody really > uses Windows because they recognise how massive a threat it is to their > privacy and security? At the moment the priority seems to be the former > and there seems to be no idea about how to approach the latter (a > "strategy"). What is it that you propose we do? Freeze the protocol so other people can implement it, even if it means we can never make a change to it again? What good would multiple barely compatible implementations do if no more people use the network? The key point is to solve the numerous fundamental problems which still exist in Freenet, to get more users, and especially to get more developers. None of that is helped by premature standardisation. Standardisation should happen after we've solved all the important problems, declared 1.0 or better 2.0, and have published rigorous proofs that everything always works. Which will probably never happen. It's much the same issue as packaging Freenet: We need a repository. We could even get into experimental/unstable. But fred cannot be in stable, because it means lots and lots of users with nodes which are several *years* out of date. I do agree that we need to be concerned for our non-Windows users. In particular, in the long run we will have to make Freenet run well on small Linux dedicated boxes (Pi's etc), because people won't have fixed computers. Of *course* we should document the protocol. We should document everything, to make life easier for new developers. But it's not a panacea.
signature.asc
Description: OpenPGP digital signature
_______________________________________________ Devl mailing list Devl@freenetproject.org https://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl
