I think this is an important point. We really need an established tagging scheme for signal phases, particularly protected turn phases, not only for more accurate driving time estimates but also for pedestrian safety. At a standard four-way intersection of undivided streets, a crossing pedestrian needs to watch for turning traffic from up to three directions simultaneously, even if they have a dedicated walk signal. The number of conflicts depends on a) whether turns are protected versus permitted, b) whether turns on red are permitted, and c) whether there are any slip lanes.
The following observations are based on norms in the U.S., but there are probably some commonalities with other regions.
This proposal only considers traffic turning right from the parallel street onto the cross street with a permissive green signal. A split phasing configuration would be tagged as shared_green=yes, which accurately conveys the fact that the pedestrian has to watch for right-turning traffic. However, unprotected left turns can be just as much of a hazard to the pedestrian, if not more, because the driver may dart into a gap in incoming traffic unaware of the pedestrian.
Even if all turns from the parallel street are protected, if traffic on the cross street can turn right on red, then there may still be some risk to the pedestrian. In theory, that conflict would be minimal, because the driver will have stopped at the stop line before proceeding to turn right. However, where I’ve lived, drivers usually pull forward to where they see oncoming cars, without first stopping before the crosswalk.
When an intersection has slip lanes, traffic turning right on green would cross a smaller crosswalk that only has to watch for traffic from one direction. If the slip lane has a dedicated signal, the crosswalk would be tagged shared_green=no. But usually the slip lane has no dedicated signal, just a yield sign at most, so the crossing would be tagged crossing:signals=no, thus shared_green=no. Ironically, slip lanes tend to encourage speeding and are quantitatively less safe to pedestrians than even the unprotected turns discussed earlier. Meanwhile, the longer crosswalk across the cross street would also be shared_green=no on account of the slip lane, even if left turns across this crosswalk are unprotected.
In conclusion, while I applaud the goal of this proposal, I’m concerned that it would mislead data consumers and lull pedestrians into a false sense of safety in some common situations. I suspect we can better communicate these hazards by tagging traffic signals with their phases, but most routers don’t yet analyze traffic flow within an intersection to any degree.