This paper presents the design and first results of the Dynamic Speed Limit (DSL) experiment that took place on the last 13 km stretch of the B-23 freeway accessing the city of Barcelona (Spain). The DSL system installed, in addition to the high density of surveillance equipment available, makes this stretch a suitable highway lab. The objective of the experiment was to construct a comprehensive database of traffic engineering variables on a freeway site when different speed limits apply. Detailed measurements of vehicle counts, speeds, occupancies, lane changing maneuvers and travel times were taken. The potential of this multi-source database is huge. For instance, preliminary analysis empirically proves that drivers’ compliance with dynamic speed limits is very limited, unless speed enforcement devices are present. In addition, in controlled sections with high compliance rates, it is observed that lowering the speed limit extends the range for critical occupancies beyond the typical values without changing the freeway capacity. This might be due to the observed reduction in the lane changing activity at these occupancy levels and for low speed limits. A direct consequence of this fact is that, even with a high compliance to very low speed limits (e.g. 40 km/h), the flow restriction imposed is insignificant. Stable flows near capacity are maintained together with very low speeds and very high densities. In practice, this means that the mainline metering capabilities of DSL are limited.