Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uwm.edu!zaphod.mps.ohio-state.edu!mips!wdl1.wdl.loral.com!spl27!pcg
From: pcg@spl27.spl.loral.com (Paul C George)
Newsgroups: comp.software-eng
Subject: Re: Effort estimation based on language
Keywords: cost estimation, effort estimation, language
Message-ID: <1991Feb1.210822.6718@wdl1.wdl.loral.com>
Date: 1 Feb 91 21:08:22 GMT
References: <1900@blackbird.afit.af.mil> <9997@orca.wv.tek.com>
Sender: root@wdl1.wdl.loral.com (SUPER USER)
Reply-To: pcg@spl27.spl.loral.com (Paul C George)
Organization: Loral Software Productivity Laboratory
Lines: 18
Nntp-Posting-Host: spl27

I believe this the Work of Ed Siedewitz (creator of GOOD method) at Nasa
Ames reasearch center. The application was flight dynamics simulators for 
spacecraft, including the control system, Attitude dynamics and environment.
This was a well understood domain where new projects were slight deltas from
previous.

A report is available from NASA, but as it is at home I can't give you the
citation. email me if you need it.

As to productivity, he attributed most of the lack of improvement in error rates
to the lack of familiarity with Ada concepts, and the fact that a soon as a 
programmer got productive, they were hired away by commercial firms. As a data point, by the 4th Ada project, they were producing 130 SLOC/day vs 28.8 for
fortran. Statements/day was 24 vs 14.4.

Reuse was 32% for the 2nd Ada project, reaching 90% by the 4th. By comparison
Fortran projects ranged from 16 - 36%

(these figures are from his presentation in the procedings of the 4/90 CASEWorld conference)