www-community mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stefano Mazzocchi <stef...@apache.org>
Subject Re: [ANN] Introducing Apache Agora - reloaded!
Date Fri, 15 Jul 2005 05:24:25 GMT
Stefano Mazzocchi wrote:
> Ian Holsman wrote:
> 
>>How would you compare it against Microsoft's Netscan
>>(http://netscan.research.microsoft.com/Static/Default.asp)
>>?
>>which also tries to find the main contributors in different communities.
> 
> 
> I think "main" implies metrics and I really didn't want to go there. I
> think contribution is inversily proportional to the distance from the
> center of gravity of the group, but I wanted to keep it subjective to
> avoid building altars than that people want to fight to step on.
> 

sorry, hit sent too soon.

>>Is 'agora' public knowledge?

no 'private' mail list is being analyzed, so yes, it's public knowledge.

it has not been largerly publicized (yet) but I wouldn't be against
putting it in a more visible position on the apache.org web site.

>>what does the 'decay' area do?

if you do one reply to a message of mine, agora creates a link between
you and me of strenght 1.0, then if you do another reply this gets
added. Note that links are directional: you might reply a lot to me, but
I never reply to you, this is still calculated in the graph drawing
algorithm.

Decay means that you get 1.0 if you reply now and exponentially lower
value if your reply was earlier in time.

I introduced this because I was curious about how much the past of a
project (especially if you load a lot of months of a project in memory)
was influencing its present.

Rather surprisingly, decay does *NOT* introduce substantial difference
in the way the graph is shaped or the position of people in the graph,
which is a very very interesting property and I have no idea why that is
the case.

>>How does one differentiate between a useful communication and a flame
>>war? 

There is no attempt to do, ATM.

>>I remember seeing Mark Smith (the netscan developer) talk about how
>>he could identify the different types via the length of the conversation.

As I mentioned earlier, we don't tend to host a lot of inflammatory
people in Apache (don't really know why, I suspect is an historical
thing or avoiding to react agressively to aggressions, which make
flamelovers go somewhere else, but I don't know how to test this
hypothesis), this keeps the signal/noise ratio high.

Identifying a conversation means that at least *you* can pretend to
understand the difference between inflammatory and not. I suspect this
difference is also very cultural: a conversation that is a 'normal' tone
in one community might be considered very 'strident' in another. I'm
sure I'm not the only one who has experienced this.

At the end of the day, I'm a big fan of the love/hate hypothesis:
replying to somebody indicates a sort of preferential attachment, no
matter what you are saying. Ignoring them is the only signal that the
communication is not useful.

NOTE: I do *not* think that the size of the social cluster is an
indication of health, there is something else that influences it... but
I don't know what it is (yet).

>>Overall a big '+1'

Thanks.

-- 
Stefano.


---------------------------------------------------------------------
To unsubscribe, e-mail: community-unsubscribe@apache.org
For additional commands, e-mail: community-help@apache.org


Mime
View raw message