Abstract-This paper aims to exploit spectrum white spaces in time-domain via the Dynamic Spectrum Access (DSA) technology. DSA relies on opportunistic access of licensed spectrum by unlicensed devices. Any "unlicensed-with-licensed" coexistence must ensure "safe" (i.e., un-interfered) communications for the incumbents, while achieving high spectrum-use efficiency for the secondary users. We propose a novel and comprehensive metric called the Coexistence Goodness Factor (CGF) to accurately model the inherent tradeoff between incumbent safety and unlicensed access efficiency for time-domain DSA-based coexistence. To optimize the coexistence performance, we propose a generic, online, dual-mode DSA coexistence protocol. The unlicensed devices attempt to estimate incumbent behavior patterns, and enter the Aggressive Mode (AM) once such a pattern is found, while they stay in the Safe Mode (SM) otherwise. For low-overhead and reliable estimation of incumbent spectrum-usage patterns, we propose algorithms based on Approximate Entropy (ApEn). Further, we design Spectrum-Conscious WiFi (SpeCWiFi), which provisions the proposed DSA coexistence scheme to the base 802.11 MAC. We conduct an extensive experimental evaluation of SpeCWiFi using a MadWifi-based prototype implementation in conjunction with 802.11 wireless cards. The evaluation shows that SpeCWiFi achieves excellent time-domain DSA coexistence in the presence of different types of licensed spectrum, including fast-varying channels that feature short-duration white spaces.