RE: BitTorrent Methodology

From: Ben Grodsky <grodsky_at_mediadefender.com>
Date: Thu, 3 May 2007 15:01:16 -0700

dude, i wasn't being sarcastic. i'm really impressed with your answers. they'll definitely be fine for at least a while. i'm sure they're still grappling with dumb details like "eDonkey user hash."
 
i agree that this is borderline proprietary questioning they have. what they should have done is talked to you about data feeds with Necip. you could have gone field by field in the db dumps you had printed out... had they not preferred to twiddle their thumbs, check e-mail, and talk about "conversions." i realize they didn't know you had those print-outs, but obviously there are things that can be done in person that are difficult to do via phone. they should have focused on things that require pointing at other things (shit, they could've had you explain field by field the stuff on the Reporting website.... that would've been a better use of your time).

side note. miivi is getting pretty stable. here are my atypical user comments:

1.
        hibernation oddities: i've been using it for a while now. it freaks out sometimes when i hibernate and wake the laptop back up, but i think you could get around that by monitoring for a windows call to shutdown or hibernate forcing the downloads to pause. the only way i can resurrect miivi after hibernation is by closing the app and restarting it. the stop and start options on the systray don't bring it back from the dead.
2.
        sources/3-download limit: the shit that i download w/ few sources is indexed in searches, but probably shouldn't be. i end up having loads of crap at the top of my queue that takes hours to connect to 1 source. then, it'll run for several hours at a slow speed (because there's only one source) and it'll stop when that source stops sharing with me. the movies i'm getting range from 50 mb --> 800 mb and all download between 5-20 kb/sec, but only when the few sources are online. would it make sense to threshold Search not to display things with fewer than X sources? i'd rather not know i can't have what does exist, than not have what i know does exist. another thing that would make this sort of Weird Content queuing more useful is the ability to have more than 3 downloads.

otherwise, i'm actually enjoying using MiiVi. for now, it's a p2p prog w/ really useful thumbnails so I don't waste a bunch of time getting crap. now i can focus what i waste my time on and get frustrated the crap i like is so unpopular that i can't get it for days ;)
________________________________

From: Jay Mairs
Sent: Thu 03-May-07 14:45
To: Ben Grodsky
Subject: RE: BitTorrent Methodology

He said a quick email exchange. I don't consider giving implementation details for each individual network quick. Actually, I consider most of it to be proprietary. If Randy wants to give a low level transparency into how we do every little thing, I'll need to be told that explicitly.

 

From: Ben Grodsky
Sent: Thursday, May 03, 2007 2:37 PM
To: Jay Mairs
Subject: RE: BitTorrent Methodology

 

man, you're good. it's almost as if you said something, but not really. :)

 

they really screwed up by not letting you and Necip duke it out scientist vs. scientist. that would've been a great miivi clip.

________________________________

From: Jay Mairs
Sent: Thu 03-May-07 14:33
To: Skinner, Andrew (NBC Universal); Ben Grodsky
Cc: Markham, Aaron (NBC Universal); Ben Ebert
Subject: RE: BitTorrent Methodology

Andrew,

 

I assume these questions are for all networks, not just Bittorrent (as the subject implies).

 

1. IP addresses are not (and cannot be) collected by protection(countermeasures) computers. In fact, this is why we developed a separate data collection system for your data feeds.

2. Supply collection implementation details vary by network. In general, supply corresponds to a user sharing a file with a certain hash on a certain day. A user will be counted again if revisited on a different day, but not if it's the same day.

3. Demand collection also varies by network. Demand is defined as a keyword search by a user. These are counted daily to give a daily demand number. If a user types in a search string and hits the search button, that will be counted once. If the same user does a second search, that will also be counted once.

4. Our countermeasures are removed prior to the calculation of the demand and supply numbers, so they don't have any effect on them.

 

Regards,

Jay Mairs

 

From: Skinner, Andrew (NBC Universal) [mailto:Andrew.Skinner_at_nbcuni.com]
Sent: Thursday, May 03, 2007 1:50 PM
To: Ben Grodsky; Jay Mairs
Cc: Markham, Aaron (NBC Universal)
Subject: RE: BitTorrent Methodology
Importance: High

 

Next Thursday will be fine. I think that most of the questions about supply and demand can be answered through some quick email exchanges though. Jay, would you mind sparing a few moments to clarify the following:

 

1. Are IP addresses time stamped and collected during countermeasures? Is it possible to figure out if countermeasures were used on user X, and If so, how many times and when? During the call we will most likely be discussing these questions in regard to ed2k user hash.

 

 

2. How is supply measured and collected? Basically, how do you define supply? If a swarm is entered and supply data is captured will existing users be counted again if it is revisited at a later time?

 

 

3. How is demand measured and collected? How do you define demand?

 

 

4. How do spoofs, decoys, swarming, and interdiction affect supply and demand numbers? Are these counts included, counted separate, or not included in the supply/demand numbers?

 

 

 

Thanks,

 

Andrew

 

 

________________________________

From: Ben Grodsky [mailto:grodsky_at_mediadefender.com]
Sent: Wednesday, May 02, 2007 12:19 PM
To: Skinner, Andrew (NBC Universal); Markham, Aaron (NBC Universal)
Cc: Randy Saaf; Jay Mairs
Subject: RE: BitTorrent Methodology

 

Aaron and Andrew,

Attached please find our BitTorrent testing methodology. This is what our QA team does. We added the last step as a way for your team to help us improve our BitTorrent effectiveness for your titles.

Thanks,

Ben

________________________________

From: Ben Grodsky
Sent: Tue 01-May-07 15:13
To: 'Andrew.Skinner_at_nbcuni.com'; Jay Mairs
Cc: 'doganaksoy_at_crd.ge.com'
Subject: Re:

Andrew,

I'm out Thursday (this week) through Wednesday (next week). Could we set a call for Thursday next week?

Thanks,
Ben

----- Original Message -----
From: Skinner, Andrew (NBC Universal) <Andrew.Skinner_at_nbcuni.com>
To: Ben Grodsky; Jay Mairs
Cc: Doganaksoy, Necip (GE, Research) <doganaksoy_at_crd.ge.com>
Sent: Tue May 01 14:56:03 2007
Subject:

Hi Ben & Jay,

Are you guys open for a call this Friday? Necip and I would like to discuss how supply and demand are measured, as well as how user hashes are collected. I'll be free for the rest of the day if one of you would like to clear some stuff up ahead of time.

Thanks,

Andrew

818.777.3491
Received on Fri Sep 14 2007 - 10:55:54 BST

This archive was generated by hypermail 2.2.0 : Sun Sep 16 2007 - 22:19:46 BST