[Wikiquality-l] Working on blame tool

[jamesmikedupont at googlemail.com]

On Sat, Oct 17, 2009 at 9:50 PM, Brian J Mingus
<Brian.Mingus at colorado.edu> wrote:
>
>
> On Sat, Oct 17, 2009 at 12:40 PM, jamesmikedupont at googlemail.com
> <jamesmikedupont at googlemail.com> wrote:
>>
>> I was not able to find any examples.
>> I think that such a blame and trust tool belongs in git, not in
>> wikipedia because there are many other usages for it.
>> mike
>>
>> On Sat, Oct 17, 2009 at 8:33 PM, John Erling Blad
>> <john.erling.blad at jeb.no> wrote:
>> > There is a student at UiO looking into alternate trust coloring schemes.
>> > John Erling /jeblad
>> >
>> > jamesmikedupont at googlemail.com wrote:
>> >> On Sat, Oct 17, 2009 at 7:39 PM, Platonides <platonides at gmail.com>
>> >> wrote:
>> >>
>> >>> jamesmikedupont at googlemail.com wrote:
>> >>>
>> >>>> FYI,
>> >>>> I am working on a blame tool for wikipedia
>> >>>>
>> >>>> http://fmtyewtk.blogspot.com/2009/10/mediawiki-git-word-level-blaming-one.html
>> >>>> thanks,
>> >>>> mike
>> >>>>
>> >>> Importing an article history into git for using git blame doesn't seem
>> >>> like a good method...
>> >>>
>> >>
>> >> Well importing it just for blame is bad. I agree. I read about the
>> >> wikiblame.
>> >>
>> >> my purpose is to port the wikipedia over to git...
>> >>
>> >> mike
>
> So far all of the implementations of blame tools for the full  history dump
> of a wiki do not have the features of an ideal blame tool.
>
> Given an arbitrary string of text an ideal blame tool can scan Wikipedia's
> entire history - and ideally the history of all WMF wikis - and tell you the
> authors of that text.
>
> The design of such a system is essentially a search engine where each
> revision is a page with an associated author. The engine works iteratively,
> first finding all page blobs (where a page blob contains all text across all
> revisions for an article) that contain all of the words being searched for,
> and then iteratively working forwards in time on the revisions of that
> article in an effort to find the earliest authors. This isn't a complete
> spec, but it gives the general idea.

Nice description.

Well imagine a problem of finding and removing some copyrighted code
from linux kernel,
or some bug from software. We need to make sure that git has these features.