The majority of linear-optical nondestructive implementations of universal quantum gates are based on single-photon resolving detectors. We propose two implementations, which are nondestructive (i.e., destroying only ancilla states) and work with conventional detectors (i.e., which do not resolve number of photons). Moreover, we analyze a recently proposed scheme of Wang et al. [J. Opt. Soc. Am. B \textbf{27}, 27 (2010)] of an optical iSWAP gate based on two ancillae in Bell's states, classical feedforward, and conventional detectors with the total probability of success equal to $\eta^4/32$, where $\eta$ is detector's efficiency. By observing that the iSWAP gate can be replaced by the controlled NOT (CNOT) gate with additional deterministic gates, we list various possible linear-optical implementations of the iSWAP gate: (i) assuming various ancilla states (unentangled, two-photon and multiphoton-entangled states) or no ancillae at all, (ii) with or without classical feedforward, (iii) destructive or nondestructive schemes, and (iv) using conventional or single-photon detectors. In particular, we show how the nondestructive iSWAP gate can be implemented with the success probability of $\eta^4/8$ assuming the same ancillae, classical feedforward, and fewer number of conventional detectors than those in the scheme of Wang {\em et al.} We discuss other schemes of the nondestructive universal gates using conventional detectors and entangled ancillae in a cluster state, Greenberger-Horne-Zeilinger and Bell's states giving the success probability of $\eta^4/4$, $\eta^6/8$, and $\eta^4/8$, respectively. In the latter scheme, we analyze how detector imperfections (dark counts in addition to finite efficiency and no photon-number resolution) and imperfect sources of ancilla states deteriorate the quantum gate operation.