According to the stats, the number of new users and the number of one-time users is about equal: http://127.0.0.1:8889/freenet:USK at gjw6StjZOZ4OAG-pqOxIp5Nk11udQZOrozD4jld42Ac,BYyqgAtc9p0JGbJ~18XU6mtO9ChnBZdf~ttCn48FV7s,AQACAAE/graphs/1238/ (The second graph; the first graph promisingly appears to show that numbers are starting to rise again) According to evan's work last year, the number of users trying freenet for 5 minutes and then uninstalling is relatively low.
However, according to Google Code and Google Analytics, we get 10,000-12,000 downloads per month. (Granted we only have stats for two months on Google Analytics for some reason). THEREFORE, a large proportion of users are dropping out before or during the post-install wizard. Possibilities: - They download it but never install it. - They cancel the install. - They never click on the rabbit. - They drop out at some point during the post-install wizard. IMHO the last is the most likely. Unfortunately we can't tell without spying on our users, which would be unethical and risky. :| Logically, they must drop out before the node is operational. That leaves: - The intro page. We could get rid of this but it's probably better with it. Isn't it? We could add an advanced/simple mode. - The browser warning. This is scary and demanding. We really need to get rid of this where possible. And it is in fact possible now thanks to Firefox 4 fixing the CSS history bug! I started to look into this yesterday. - Updating and plugins. Some users will want to configure autoupdate and will be offended if they're not asked, but arguably this should be an advanced mode setup setting... Plugins - jargonish. UPnP - jargonish, but most users can answer it. STUN - jargonish and scary, but generally necessary for darknet! (It may be possible to get away without it with enough FOAFs). These settings should probably only be shown on specific security levels. - Opennet/darknet page: We're going to lose a lot of users here... Fundamentally Freenet is insecure in opennet mode, but everyone wants to use it in opennet mode. And rather than our historical strategy of vaguely hinting, we are fairly direct about this now. We also have a detailed explanation in a mouseover, which is a bit ugly; maybe it should be a separate page. IMHO better darknet support/invites etc could reduce the drop out rate considerably and could get us a lot more users. - LOW vs NORMAL: again this mentions that opennet is low security, but isn't too jargony... By this point, the node is operational. However, bootstrapping can be very slow (which is another problem which is largely opennet-specific; on darknet the corresponding issue is even with FOAFs there may not be enough peers). - Physical security: Maybe excessively complex? Perhaps we could default to NORMAL, and offer to set a password after the user has queued some stuff? I'm not sure ... Scary language here too, but at least it's a choice... It might be possible to simplify it, i.e. don't tell the user as much until after a choice is made. - Bandwidth limits: Many users don't know their upload bandwidth limits. It is difficult to do anything about this though... - Datastore size: This is straightforward, but if we try to skip it many users (and not just geeks) may get angry about it. - Congratulations page: Possibly redundant, worth considering. IMHO: - It is worth doing some work on the browser warning page. It should not generally be shown at all, if the browser is Firefox 4 or later and privacy mode is enabled. The caveats here are privacy mode may not actually work on recent FF, it just opens another tab; we need to file a bug for this ... We may want to show an extra warning page later on if the user chooses higher physical security levels. IMHO this should gain us a significant percentage. - Updating should be on by default. This is a strong candidate for making dependant on advanced mode setup. Which could also unconditionally enable the next lot of questions ... - UPnP we should only ask if network seclevel is above LOW. Basically it's "are you on a student LAN?", although eastern europeans need to turn it off too. "Are you on a local network including people you don't trust? (I.e. not family etc)" ??? Arguably UPnP isn't all that dangerous anyway, the worst case is somebody intercepts our incoming traffic, they won't be able to do much with it and we wouldn't have been able to communicate at all otherwise - they'd need to conspire with a corrupt seednode as well so it gets rather unlikely... TODO: Ask nextgens exactly what the consequences of evil UPnP are for our purposes. - STUN is a pig. We should ask the user if they have a higher security setting. However, if they turn off STUN, we have no way to get their IP address, apart from UPnP or them setting one specifically. We shouldn't ask for a static IP unless we're desperate or they are in advanced mode. Darknet on a student LAN is potentially very problematic. On the other hand if you have a darknet invite including some peers who are port forwarded, it may be workable. TODO URGENT: Release a new build of JSTUN with newer server list. - In both of these cases, we need to ask the user when they upgrade to a higher security level - even if they don't initially set a high security level. - The opennet/darknet dilemma can only be fixed by darknet invites IMHO. Which will change it slightly - if you already have friends this should certainly have an impact. And whether you have connected to them already might even have an impact on whether you need JSTUN or UPnP, it could be postponed to a useralert later on??? - We could get rid of the LOW/NORMAL choice (but not HIGH/MAXIMUM) if the user is not in advanced mode. -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 198 bytes Desc: This is a digitally signed message part. URL: <https://emu.freenetproject.org/pipermail/devl/attachments/20110714/9ea48ddf/attachment.pgp>
