foot profile routes on highway=construction

[emiltin]

this is because profile.avoid does not include 'construction', and no tests cover this.

[maraf24]

It may be intentional. For example in urban areas pedestrians are often allowed to use such roads to get to the destination i.e. shops or houses.

[Lee-Carre]

This issue, I think, is the one I'm encountering with Moneypenny Lane.

@maraf24 makes a good point which I hadn't considered. However, other algorithms (given the same data & origin + destination) do route around the lane (usually via Vauxhall Street then the footpath (way #938088198, way #938088199, way #173615145)).

In this case, the entire road (including sidewalk=*) is new. No traffic (except construction) is allowed to traverse it.

I presume access=no (or at least foot=no) should make the situation clear to OSRM? Recently set access=no in order to check the results, but I gather that'll take ~24 hours.

If so, the question then becomes one of how to handle cases where access-condition tags aren't explicitly set.

Having perused other issues in search of this one, a common response to complaints about routes avoiding any possible construction is that this is the safer option. However, a counterpoint for on-foot navigation is that a shorter route is much preferable.

Maybe these sorts of (ambiguous) situations need to have flags which can be toggled in order to influence the algorithm as a user preference for default behaviour for their locale?