We use type Ia supernovae (SN Ia) data in combination with recent baryonic acoustic oscillations (BAO) and cosmic microwave background (CMB) observations to constrain a kink-like parametrization of the deceleration parameter ($q$). This $q$-parametri
zation can be written in terms of the initial ($q_i$) and present ($q_0$) values of the deceleration parameter, the redshift of the cosmic transition from deceleration to acceleration ($z_t$) and the redshift width of such transition ($tau$). By assuming a flat space geometry, $q_i=1/2$ and adopting a likelihood approach to deal with the SN Ia data we obtain, at the 68% confidence level (C.L.), that: $z_t=0.56^{+0.13}_{-0.10}$, $tau=0.47^{+0.16}_{-0.20}$ and $q_0=-0.31^{+0.11}_{-0.11}$ when we combine BAO/CMB observations with SN Ia data processed with the MLCS2k2 light-curve fitter. When in this combination we use the SALT2 fitter we get instead, at the same C.L.: $z_t=0.64^{+0.13}_{-0.07}$, $tau=0.36^{+0.11}_{-0.17}$ and $q_0=-0.53^{+0.17}_{-0.13}$. Our results indicate, with a quite general and model independent approach, that MLCS2k2 favors Dvali-Gabadadze-Porrati-like cosmological models, while SALT2 favors $Lambda$CDM-like ones. Progress in determining the transition redshift and/or the present value of the deceleration parameter depends crucially on solving the issue of the difference obtained when using these two light-curve fitters.
