1 |
On Fri, Mar 27, 2009 at 10:09 AM, Joachim Bartosik <jbartosik@×××××.com> wrote: |
2 |
> Today I talked to my friend about the project and a problem I have with |
3 |
> stats severer/ client ( I'm still not satisfied with those I wrote about). |
4 |
> He is participating in project which is about trust management so the talk |
5 |
> was very valuable. I got a list of texts I should talk a look at unluckily |
6 |
> most of that requires peers reviewing peers and I think it would be hard to |
7 |
> apply to this project ( I plan to continue reading though just in case there |
8 |
> is something valuable for this project). Here are conclusions of this talk: |
9 |
> * users are really lazy - if we ask them to type their emails and click a |
10 |
> link many will be too lazy to do that. |
11 |
> * email-registration can be fooled. |
12 |
> ( well I knew those two things but I didn't have any better ideas so I |
13 |
> sticked to it) |
14 |
> * we can assign level of trust to each host and when creating statistics |
15 |
> simply ignore those with low-level trust ( like we trust hosts that are |
16 |
> submitting data as asked for a long time more then new comers). |
17 |
> * we can make registration resource-consuming ( like solve some NP-complete |
18 |
> problem) not enough to make users angry ( run on low priority for a few |
19 |
> minutes on an average Gentoo user computer) but enough to make life of |
20 |
> people who would like to put a lot of fake data in our DB harder. |
21 |
|
22 |
I've been super busy at work, but I'll try to think about this and get |
23 |
back to you tonight or tomorrow. |
24 |
|
25 |
-A |
26 |
|
27 |
> |
28 |
> If we combine those two it would be probably both easier for real users and |
29 |
> provide better quality of data. What do you think? |
30 |
> -- |
31 |
> Joachim Filip |
32 |
> |