On 2013/08/05 (Aug), at 5:28 PM, Matthew Toseland wrote: >> There might be a stat we can collect in this respect.... IMO, we should have >> a timer that fires if a node has not been initialized after N days (1 day? >> 2?) that makes a HTTP request to the server saying, "I'm running, but >> lost"... maybe reporting how many (if any) wizard screens they got through, >> esp number of FProxy requests (did they even hit the first screen!). > > If it was downloaded via Tor or via filesharing etc then it would need to > know that so that it could not report in. IMHO we need explicit informed > consent here.
While I certainly understand your perspective (maybe even that of our "target audience"), I disagree. What we are trying to do here is diagnose an unknown failure case, and it is already the case that our "target audience" (that would appreciate the app not phoning home) can generally get Freenet up & running. IMO, it is enough that: * the delay be long enough that even a "mild geek" would get it working before such a time (or the time to setup a darknet connection?), * the "most personal" info transmitted be IP address (unavoidable) and System type (which platform is broken?) * the transport be HTTPS To add explicit consent only adds an additional step they might get lost at (or balk at). > Maybe a checkbox at the beginning of the installation process - before even > installing Java? > > Only a handful of users will tick it, but stats from them would be useful. > > [ x ] Tell the Freenet Project how the installation went (warning this will > make an encrypted connection to our server freenetproject.org). How about avoiding the scary words, default to true (they already d/l it from our server, right?) thusly: [x] Report installation failures over HTTPS > Or better ... have a 10% chance of asking the user explicitly: > > Do you want to help us to improve Freenet? If you click yes, the Freenet > installer will tell us whether your install of Freenet succeeded, and if it > failed then at what point it failed. No other information is sent to us, but > as a bunch of geeks who mostly use Linux, this is very useful information! > Only click Yes if you downloaded Freenet from our website. Which (to the user) roughly translates to: Would you like to help us improve Freenet? If you click yes, [.... ugghh, a paragraph?? I have to read? How do I get to the pictures??...] > There are lots of possibilites: > - Ran away before running the installer. Due to AV warnings or lack of > signature. > - User runs away when realises need to install Java. > - User runs away while installing Java (it can take a while and doesn't > always show much on screen, can assume it's borked and cancel). > - Problems installing Java. I didn't think about this one.... I wonder if there are common corporate policies against installing java. > - Problems installing Freenet. > - Didn't see the rabbit icon / didn't try it. > - Didn't complete the wizard - at what stage did they drop out? > > How can we distinguish between them? AFAICS a feedback mode is the obvious > way to do this. I agree.... unless you mean a "please tell us why your leaving the installer" sorta feedback system. ... but it sounds like we might need two nearly identical mechanisms. One for installation failure, another for installed-but-not-initialized. -- Robert Hailey
