We present formulations of two condition measures (one for linear programmi
ng (LP) due to Ye, and the other for a matrix) as optimization problems ove
r sign constraints. We construct, based on LP duality, a dual characterizat
ion of Ye's condition measure in the setting of Karmarkar's form. The eleme
ntary formulations (utilizing the dual characterization) lead to trivial pr
oofs of some results relating these two condition measures for polyhedra in
Karmarkar's form. Such interpretations, using the sign constraints, allow
for the definition of families of condition measures that are "between" the
two condition measures. Our viewpoint provides further understanding of th
e relationship of the two condition measures. As a result of this new under
standing, we point to a connection with oriented matroids and prove that a
conjecture of Vavasis and Ye on the relationship of these two condition mea
sures is false. (C) 1999 Elsevier Science B.V. All rights reserved.