[Page Update] Junction Style Guide

[sketch]

The segment before the off ramp, and the segment after on ramp (the last and first segments of the highway on either side of the detour) must have IDENTICAL names. This means exact spelling including cardinals, spaces, etc., AND city names.

The thru segments between the ramps must also have identical names. To prevent a detour, there must be a non-detour.

A lone segment is not considered a detour—has to be two or more. Where this would actually apply (given an otherwise properly edited map) I don't really know.

[sketch]

I do still want to add images, but for the currently-listed examples instead. Hence, I left the box, though I should've changed its contents.

So just to clarify, we are no longer doing wayfinders for single lane drops on the right side where the approaching signs clearly indicate the right lane is for exit only unless the continuation path has more of an apparent curve than the exit path?

Right, and we still use them where there are 3 travel lanes before the exit, one lane continues, one lane is dropped, and the lane between them is an "option" lane, which can do either. This is under the second rule:

The continuation path on a controlled-access highway or freeway has the same amount or fewer lanes as the exit path after the split;

In the future, I want to clarify and adopt some terminology which will make that clearer, but I haven't yet had time to do that.

[sketch]

To me, this is perhaps one of those "where signage indicates" situations or something like that. You're choosing which way to go on I-5...

But in this particular situation, I could go either way. You are, after all, staying on SR-99 S, and it is the "natural" continuation of the road – not just because of the lane configuration, but also on a more general scale, you're still going southbound. That said, you're entering a different road, and I think that is one criterion that sways in favor of a wayfinder. I generally agree with CBenson – err on the side of too much. But here, it is the obvious continuation and you are staying on 99. So, I dunno, judgment call. Maybe ask jemay his reasons.

[sketch]

I agree completely! But to be told, effectively, that you'll go wrong if you don't change lanes is also misleading.

Without the "continue" instruction we seem only to have "least bad" solutions to this

It says no such thing.

"Stay to the left". Remain on the left side of the road. Don't go to the right side of the road, just stay where you are, on the left. Unless you aren't already on the left, then make sure you stay to the left side. Left being the side on which you stay; and the side on which you are staying shall be left. Right, shalt thou not stay, unless thou then proceedst to the left. You know the rest

[sketch]

"Stay to the left", not keep left, just to be clear — at least, using TTS voices, and if you aren't using TTS, well...

Anyway, you are staying, to the left, where the road splits. At any split, whether an exit or whatever, both sides are roughly more or less straight, compared with a turn, anyway. So you're just staying to one side or the other of the gore point. That's the idea.

[sketch]

But this isn't just any old freeway continuation, it's a point at which you join another freeway. Geometry isn't the only consideration when determining whether a "stay to the [x]" might be helpful. For one, IMO, it should be the default at the end of any freeway, regardless of whether the physical configuration strongly suggests one direction over the other. In some situations, this can be deviated from, for example if a particular Interstate designation ends, but some historical freeway designation continues — for example, the Pontchartrain Expwy in New Orleans, which is concurrent with parts of I-10 and US-90 BUS. US-90 BUS ends part way down but I-10 W is the obvious continuation, because that is the continuation of the Pontchartrain Expwy, which you are still on. (But even then, I have a "stay to the left to I-10 W / Baton Rouge" there, because of a lane drop.) The same can be seen with several of the historical freeway designations in Detroit, for instance the Chrysler Fwy, which starts with about a mile of I-375 before the I-75 designation takes over.

For this particular situation, again, I am on the fence. But since you are joining with a numbered highway of a higher stature – on a state highway, joining an Interstate highway – it's possible this is such a situation.

It is certainly valid to want to know, and to think that drivers might want to know, the designation of the highway on which they're driving. That is hardly, hardly a suggestion that Waze should speak to you at every interchange. Usually you stay on the same road and only when you leave that road do you get an instruction. My first paragraph follows the same principle. The reason I am on the fence about this particular situation is that you are staying on SR-99 S. At the same time, you are joining I-5 S.

[sketch]

AGCs are my preferred method. Maybe it's not in the list but it's something I've been doing since before the bowtie "existed".

[sketch]

For the record (getting to be a broken record I'm afraid): I disagree completely with the idea that drivers on the whole would prefer to be notified of changes in road numbering/naming even if it means getting instructions that may seem inconsistent or misleading in the immediate physical context. Personally I am happy to look at the signs to figure out that the highway is changing name/number, I want Waze to stick to telling me where to go, and I don't want to have to ask with every "do this with the steering wheel" instruction if it's for real or if Waze is just kidding. But I accept that I seem to be in the minority in this perspective

Here is an example of what can happen with an informational "stay" instruction where no physical action is necessary.

(I fixed it, but) that isn't an informational "stay" instruction, it was just a poorly-constructed interchange. It's the ramp-type continuity in the absence of name continuity that made that happen. I do not believe there was any intention to create a "stay left to I-980" instruction here, simply a thoughtless edit.

(For those reading in the future, the first I-980 E segment was formerly a ramp segment with an exit name on it.)

[sketch]

I third the motion. First option.

[sketch]

OK guys, I think I figured this out. I'm normally approaching the ramps at 70-80 mph. The other night I was stuck behind a truck doing 50, and sure enough the last exit call-out was too early. So it appears the Waze is not doing it's sums correctly when calculating the timing for the announcement. But for now, I think the classic guidance of first geo node at the gore point is still correct. You really, really don't want a late ramp call at 75 mph. Early is still better.

Then next step is to make Waze aware of the problem and see if they can fix the timing. To solve the second problem of the next command appearing too soon, I'd like to formally request that Waze delay the switch on ramps.

based on anecdotal previous testing by sketch, the speed assessment for prompt timing is not very robust. It calculates at a certain point as you approach the exit, what your speed is, and considering known or anticipated segment speeds ahead determines the appropriate times to deliver the prompts based on that. If you then change your speed after the calculations are performed, the timing does not adapt. This means if you were doing 75mph when the calculation was made, and then you slowed down, the prompt will come too early. If you speed up after the calculations, the prompt will come too late - i.e. you may miss the prompt as it may be overridden by the next prompt.

I would think asking waze to delay the switch on ramps is dangerous, as it runs the risk of no prompt being given in time when you are in a confusing interchange with successive ramps. This could easily lead users to get on the wrong highways and get delayed and very frustrated.

The reason for the "first geometry node at the gore, and junction node a bit earlier" is that the TTS/voice prompt engine evaluates once per second, and it is reactive. Once per second, it thinks, "Are we there yet?" where "there" is the calculated distance from the next junction node at which the prompt should be given. The answer is either "Not there yet, no prompt yet..." or "Oop, we're here, better say something."

So, you have a one-second window during which you reach (and pass) the point where Waze thinks you had better be instructed. Assuming that Waze's calculations are correct and just, the way it currently works, that window starts at the instruction distance (let's call it "X") and continues for whatever distance you cover during the following second (let's call it "error"). If you have the junction node set at the theoretical gore (the "decision point"), then the distance window for that prompt is from X to X – error. In other words, the best you can do is to have the prompt exactly where Waze thinks you should have it, and the worst you can do is some 30–35 meters too late, at freeway speeds.

So, setting the junction node just a bit ahead—a few dozen meters—moves that window ahead to err on the side of too early, rather than too late, which is a safer bet. This is not to mention the time it takes for the voice to actually start talking and to finish the words "Exit right ...", and while that is presumably factored into the instruction distance X, sometimes phones do lag a bit with background processes (e.g., Spotify in the car) and such.

Hence, the standard as written in the wiki.