The figures are available in color in electronic form.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

WW computed the stellar yields for stars in the 11 to 40 $M_{\odot }$ mass range. For masses larger than 40 $M_{\odot }$ we extrapolated their stellar yields, as shown by the triangles in Figs. 1 and 3. In particular, for carbon, our extrapolated value at 70 $M_{\odot }$ was chosen to match the value computed for this mass by Nomoto et al. (1997 - these latter are essentially the same as in Thielemann et al. 1996, but for an enlarged grid of masses - see open triangle and open circle in Fig. 1 for a 70 $M_{\odot }$ ). However, as concluded in CMR2003, although the yields of Thielemann et al. (1996) ensure a good agreement between data and model predictions for many abundance ratios, for carbon it was necessary to increase the stellar yields, for m>40 $M_{\odot }$ , by a factor of three. In fact, Thielemann et al. (1996) do not account for mass loss which, during the WC phase, is responsible for ejecting helium and carbon into the interstellar medium. Given this fact, in the present work we also increased the WW yields (which also do not account for mass loss) by the same factor (see filled triangles in the upper panel of Fig. 1).

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... IMF

In CRM2003 we suggest that to be able to fit the N/O ratio observed in BCGs a flatter IMF was necessary, but Lanfranchi & Matteucci (2003) concluded that such models would predict [O/Fe] ratios larger than the observed ones.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... DLAs

This value is based essentially on O/Zn and Si/Zn abundance ratios. In fact, part of Fe in DLAs is likely to be in dust form, while O and Zn are not depleted in dust. For Si only a mild depletion is expected (see discussions in Centurion et al. 2003a,b; Pettini et al. 2002; Vladilo 2002).

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.