[sldev] Your Feedback Wanted on Search Flagging !

[Argent Stonecutter]

On 2008-05-03, at 18:49, Callum Lerwick wrote:
>> * Flagged posts are handled in order of some function based on how
>> old they are and how many flags there are and what type there are.
>> For example, spam flags get an exponential function of the number of
>> flags, say 0.1 * 2^N.
>
> Or you could just sort based on number of flaggings.

I'm assuming limited resources at Linden Labs, so I'm not assuming  
that they will EVER get to all flaggings. That is, this is triage.

The function used would be Linden Labs' business, and tuned to fit  
the workload.

> Why does it need to be so complex?

It doesn't need to be any more complex than Linden Labs chooses to  
make it, but it's quite likely that it will be simpler to use a more  
complex function... for example, if it's easy to report spam then  
there's some small number below which you should completely ignore  
spam reports, because spam is a function of volume.

>> * If something is obviously being flagged falsely, the IP address of
>> the all flaggers are put on a silent blacklist for one day, and all
>> flags from these address are silently voided.
>> * The length of the blacklist is increased each time it's invoked.
>
> These don't contradict anything I've said. The details of  
> punishment are
> orthogonal to the workflow, it is to me a well covered and thus
> un-interesting problem space.

This is not "punishment", because it's silent, there's no feedback to  
the person that their flagging has been ignored. This is part of the  
triage, eliminating known sources of bad information allowing them to  
more easily locate items that are likely to need response.