Subject: Re: Need quotes for an article...
From: "Karsten M. Self" <kmself@ix.netcom.com>
Date: Thu, 23 Dec 2004 18:26:58 -0800
Thu, 23 Dec 2004 18:26:58 -0800
Please note reply-to set to linux-elitists.

on Thu, Dec 23, 2004 at 03:42:39PM -0500, Peter Wayner (pcw2@flyzone.com) wrote:
> 
> On Dec 23, 2004, at 11:10 AM, Robin 'Roblimo' Miller wrote:
> 
> >Here's the opening paragraph of an article I'm writing:
> >
> >According to MIT's <a  
> >href="Technology">http://www.technologyreview.com/index.asp">Technology  
> >Review</a>, the world's top R&D spender isn't Microsoft. It's Ford.  
> >Pfizer is number two, DaimlerChrysler is number three, and Toyota is  
> >number four. IBM is number eight. Microsoft is number twelve. While  
> >it's unlikely that all the open source software development and  
> >research in the world represents nearly as much R&D investment as  the  
> >bottom entry on Technology Review's <a  
> >href="http://www.technologyreview.com/articles/04/12/ 
> >scorecard1204.asp?p=5">"Top 15" list</a> (Volkwagen), it probably  
> >would make their (pdf link) <a  
> >href="http://www.technologyreview.com/articles/04/12/ 
> >scorecard21204.pdf">list of 150 top research spenders</a>. But how do  
> >we meaure the value of open source development in financial terms?  
> >Should we even try to measure it this way or look for a different  
> >metric that includes social value, not just money?
> >
> >
> >Anyone want to answer these questions? This is for publication....
> >
> Note also a more methodical study:
> 
> http://www.dwheeler.com/sloc/

I turned up the Wheeler estimate as well.

There's a Register story from 1998 citing Eric Green estimating core
GNU/Linux development (kernel + utilities) at $1,000,000,000, and
$4,000,000,000 for a then-typical distro CD bundle:

    The true costs of GNU/Linux development
    http://www.theregister.co.uk/1998/10/21/the true costs of linux/

His 10,000 man years at $1b works out to $100k/man-year.  Current
outsourcing may give lower rates....


Wheeler updated his estimates with the 2.6 kernel, coming up with
development costs of $612m (kernel only).

    http://www.dwheeler.com/essays/linux-kernel-cost.html


An estimate for Debian Potato (2.2) inspired by Wheeler's work suggests
a cost of $1.9b for the 2,800 packages it included:

    http://people.debian.org/~jgb/debian-counting/

The author is working on a similar study for Woody, I'm CC'ing him, hope
he follows up with roblimo:

    jgb@gsyc.escet.urjc.es

Assuming uniform SLOC/pkg and scaling from 2.8k => 17.7k packages
suggests $12b development costs for current Debian testing/unstable.
That's a 2-minute guestimate with no confidence interval ;-)


Peace.

-- 
Karsten M. Self <kmself@ix.netcom.com>        http://kmself.home.netcom.com/
 What Part of "Gestalt" don't you understand?
    America Trans Air (ATA) customer service sucks:
     http://kmself.home.netcom.com/Rants/ata-sucks.html


["application/pgp-signature" not shown]