From: doug s. <hig...@ho...> - 2014-08-22 19:25:09
|
BigData> user tracking Q. could/should we do user tracking? If so, how? What info would we collect? Where would we put it? What would/could we use the info for? -Doug more.. When I develop for freewrl, I develop to make the kind of web3d viewer that I personally like and can use for my projects. But We know little about who downloads freewrl. Or what they use it for. To let us know, they could subscribe to a mailing list, and write an essay on their experience with freewrl for our enlightenment. But they could be busy people, looking for a quick tool to show them something, and not want to get dragged into long survey response questionairres or write-in essays. Having said that, some may not mind the idea of us knowing how they use freewrl, especially if that knowledge would go toward improving the product: adding missing node types, improving frame rates, improving error messages, improving UI etc. One idea is to add some automated statistical summarization of use, and send it 'home', if the user approves. WHAT WE COULD TRACK a) scene statistics, including what tool likely authored the scene, wrl vs x3d b) hardware performance statistics c) usage quantity and pattern statistics d) GUI / UI statistics e) navigation / picksensor statistics We already do a scene summary of a few things. Hit '/' to see the node count or '.' a few times to see frame load, which indirectly shows the class of hardware its running on. Knowing how many times a particular scene is run might also add some insight - do users play the same scene over and over? Is freewrl a frequently used program, or 'tried once for fun and forgotten', or is it loaded onto school computers by robots and never actually used? Are there a few scenes on the internet that get all the attention? Does freewrl complement some other tools, for example do Blender.org modelers use freewrl to check the real-time renderability of their scene? Are scenes small or big? Do authors optimize their scenes -small textures, DEF/USE etc? What kind of un-planned stress is freewrl under? Is it missing critical nodes? Erroring? What are the projected usage trends in the future - is there some aspect that's growing quickly or suddenly? What are the seasonal trends in usage? HOW WE COULD REPORT IT Not sure. I think one way is to request a url from a cooperative website, and put the info after a ? ie restful. In that case it's always an http GET, so freewrl would only need internet client privileges. But can enough stats be packed into a restful url? Or is there a better way? WHERE WE COULD REPORT IT Don't know. I think there needs to be a server-side script running somewhere, and it would gather the raw info and store it, perhaps in an unstructured hadoop databse. Then separately big-data mining could be done to gain insights. |