Non-connected related routes - definition discussion

[si404]

If we want to have some sense of what routes are "related" to each other when not in the same connected route, I think we need to augment the data with information about what we mean by that.

So what do we mean by that?

Obviously there's a few reasons why routes that are part of the same single route, but cannot be connected as one route due to discontinuity caused by:

• taking a ferry
• crossing over a border where its a differently numbered route
• sections are not yet constructed
• cases like the E68 at the Romanian/Hungarian border - same route, lack of international communication as to where it should run (or rather, Romania's dislike of signing E roads on A roads)

And then we have stuff like I-35 with W/E routings and same-numbered (no directional suffix) spurs/loops (eg the E55 is signed both ways around Berlin). The route isn't quite the same single route, but they clearly are one route. Likewise bannered routes - it's very clear that they are part of the same route as the vanilla highway.

Then there's more debatable stuff...

Various places give routes a number, but use different shields/prefixes to reflect to standard/status - eg Australia M/A/N/B/C1, Poland A2/S2/DK2, Ireland M3/N3, or Israel F4/H4. I'd say these are related routes.

There's the question of whether parent-child relationships would be related enough - Interstates, sure, but US routes? There's also, say, parent/child stuff in France - N1xxx is typically future Nxxx (with zero-padding), but sometimes it's former Nxxx, and D Roads sometimes are similar. There's a bit of a can of worms here.

Further to that N1xxx - the former Routes Nationales in France are obviously meant to be a continuous route but different departments' different methods of numbering means they change numbers a lot, but they are all numbered related to the original route D1xxx, D6xx, D9xx, etc.

And, perhaps at the extreme end, are collections of routes that don't share numbers or names - eg the routes in Yellowstone NP, or the Mourne Coastal Route in N Ireland + the 'scenic loops' off it.

[michih]

For me, all routes of the same system which are signed the same way (shield + numbering) should count as ONE ROUTE for stats.

Obviously there's a few reasons why routes that are part of the same single route, but cannot be connected as one route due to discontinuity caused by:

• taking a ferry
• crossing over a border where its a differently numbered route
• sections are not yet constructed
• cases like the E68 at the Romanian/Hungarian border - same route, lack of international communication as to where it should run (or rather, Romania's dislike of signing E roads on A roads)

These are "interrupted connected routes" and should count as ONE ROUTE for stats.

And then we have stuff like I-35 with W/E routings and same-numbered (no directional suffix) spurs/loops (eg the E55 is signed both ways around Berlin). The route isn't quite the same single route, but they clearly are one route. Likewise bannered routes - it's very clear that they are part of the same route as the vanilla highway

I've not checked I-35 E/W but I guess it should also be count as ONE ROUTE for stats.. E55 should also be ONE ROUTE for stats because it is only split for technical reason by us.

And I'd just like to have the possibility to open HB for all segments of THE ROUTE.

It is relevant for:

• Stats
• HB
• Clinched / Traveled shield pages

HB: add direct link to interrupted connected routes
Shield page should only show one shield if there is more than one route
Processing interrupted connected routes

[si404]

For me, all routes of the same system which are signed the same way (shield + numbering) should count as ONE ROUTE for stats.

How about the E31 in Italy vs the E31 in Netherlands-Germany? The former is officially E33, but the signs didn't change when these two DIFFERENT routes (pointlessly) swapped numbers due to grid anal-retentiveness. Or the two different E401s - the official East-West one in Brittany and Normandy, and the unofficial North-South one between Amiens and Paris created because the owners of the concession of the A16 wanted to sign it as an E road?

And the exemplar examples that are why the 'related routes' feature is broken - the 5 pairs of duplicate numbered Interstates: 76, 84, 86, 87 and 88 - these are not related routes, but routes that happen to be numbered the same due to the vagaries of the system.

The 7 (and one future) I-295s are very clearly DIFFERENT routes (Jacksonville belt, future Fayetteville belt, Richmond-Petersburg bypass, spur into Anacostia, Philadelphia bypass+Trenton belt, Bronx-Queens link, Providence belt, Portland-Augusta free route) and the numbering system encourages such duplicate numbering with 3dis.

And this is before we get to Great Britain, with it's rather a lot of accidental duplicates due to poor management of road lists: A479, A594, A1042, A1056, A1114, A1199, A4102, A5028, A5152, A5271, B198, B454, B1140, B1172, B3153, B3206, B3330, B3440, B4065, B4082, B4100, B4104, B4118, B4587, B5192, B5207, B5210, B5320, B5444, B5477, B6322, B6374, B6452, B6481. And it would be remiss of me to not point out that Northern Ireland has 3 dating back to the original numbering lists nearly 100 years ago - while the B52 and B53 are probably clerical errors (there's no B51 or B54), they were obvious errors when you see the published list, and that doesn't explain the two A37s!

[michih]

Thanks! All examples are clearly ONE ROUTE for me. Same shield and same numbering.

I have clinched the whole A27 in NI! I have clinched the whole I76 in USA! Yep. I can only claim this when I really traveled all these segments. Period

[Jim]

I mostly echo @si404's points here.

There are multiple ideas intertwined here. 

What should be a connected route for the purposes of claiming a 100% clinch? 

What should be a connectedsingle route for the purposes of multi-region .list entries?

What routes should be considered related routes?

Sorry to rehash the same examples, but I think they do help.  Especially the duplicate I-76/84/86/87/88 and the duplicate number 3dis, to me these are absolutely distinct routes and should have no more relation to each other than other separate highways in usai with different numbers.  But then the routes that are currently not connected in TM but are a single route in some meaningful sense, like those interrupted by ferry crossings, incomplete construction (the I-69s), and I'd argue cases like US 20 across Yellowstone, I'd like to see us find a way to call those things single routes.  This could be either by a generalization of the "connected" idea or some new concept/term.  But we do need to keep in mind that we have some assumptions built into both the data processing and now more than ever the web side that a connected route is in fact connected with matching points at region boundaries.  Not that we couldn't break it by allowing datacheck FPs and handling those cases in the code, but it's not as simple as slapping all of the US 20 segments into a single connected route at this point.  And even that doesn't handle the I-35E/W cases that I'd like to see count as part of a single I-35 "unit" (similarly with US 11, US 25, US 70...).  CHM, if I recall, just picked one of the branches to count as mainline and treated the other as a 3di/3dus.  We elected to break I-35 into three distinct units.

Then there's what should be related routes?   You could make the case that 3dis and all usaib routes are related to their parent routes, but I-184 would be related to western I-84 only and I-384 and I-684 to the eastern I-84 only.

[oscar]

Not sure if this is relevant to this particular "related routes" discussion, but the "related routes" feature still in the HB helps users find routes that are somehow connected to the one displayed in the HB, one of which the user might've wanted to display from the outset. For example, a user who likes me puts in the URL what is expected to be the right route file, but actually isn't (for example, r=qc.qc138, which takes you to the longest segment rather than the one in Montreal). Hitting "related routes" will guide the user to the right file.

That use of the "related routes" feature seems to work fine for that purpose, but not for other purposes.

[yakra]

We had discussion elsewhere about the flaws of rte=, but my quick summary:

• It relates routes that aren't related, like the unrelated routes worldwide that have names like "A4",

Restricting mapview's parameters by system and/or region can get us around this, as done in the stopgap fix for system.php mapview links.

• It fails to relate routes that should be related, like the I-35E/W pairs, and 3dis

The old mapview (still running on lab2) with its regex-based rte= functionality handled I-35 E/W. Not 3DIs of course.

the unrelated U.S. Interstates I-84, I-86, I-87, and I-88

These would still be displayed in either case.


This whole topic strikes me as a solution in search of a problem; an effort to box users into one particular way (just what particular way TBD?) of thinking about these routes, one that might not be appropriate or relevant to all users. This thread itself is evidence there are different ways of thinking about it. We have michih's examples upthread, and Jim & si404's counterpoints to these same examples. We have same-numbered, similarly-numbered, bannered/child routes, 3DIs etc., and probably as many opinions on what should & shouldn't count as "related".
IMO this whole idea should just be left alone.

[michih]

This whole topic strikes me as a solution in search of a problem

Disagree!

[yakra]

But it does strike me as a solution is search of a problem!
HA! Don't mind me, I'm just being pedantic in my use of language for the sake of a laugh.

Seriously though, I stand by my statement in my last post -- there are different criteria for what may or may not be "related", and these criteria will be interpreted differently by different users. (For example, should 3-digit Interstates be considered "related" to their parent Interstates?) We're practically bound to not reach consensus on these questions. Boxing users unto a particular way of thinking about related routes would IMO be a disservice to the community at large.

@michih, IIUC it sounds like, the way you envision related routes as described upthread, they're routes with the same designation, the same Route+Banner combo. Clinch I-76 west & I-76 east to clinch I-76, etc.
To search & display routes in this regard, we wouldn't need to change anything about how info is stored in the DB. We'd just need to tweak mapview to filter by banner...

[si404]

Not sure if this is relevant to this particular "related routes" discussion, but the "related routes" feature still in the HB helps users find routes that are somehow connected to the one displayed in the HB, one of which the user might've wanted to display from the outset. For example, a user who likes me puts in the URL what is expected to be the right route file, but actually isn't (for example, r=qc.qc138, which takes you to the longest segment rather than the one in Montreal). Hitting "related routes" will guide the user to the right file.

Yes, this is why the feature is useful and should continue to be provided - the issue is the unrelated routes dragged in (and the related routes excluded by a filter).

the unrelated U.S. Interstates I-84, I-86, I-87, and I-88

These would still be displayed in either case.

Jim's point is that they shouldn't be!

IMO this whole idea should just be left alone.

And then we don't get the feature back as we don't come up with a consensus of what it's there for...

And here's what I think of the filtering by system:

The way around that would have been to put a sys= filter into the link, as was temporarily done in the links from system.php:
https://travelmapping.net/user/mapview.php?rte=A1&sys=deua
...But we never did that for the Related Routes links, and now, for better or worse, the feature is gone.

That doesn't deal with:
1) bannered routes being related routes - a system filter is too narrow.
2) Jim's main gripe that duplicate numbers (eg I-88) are not related routes - a system filter is too broad.

It's a misdiagnose of the problem to just whack a system filter on it. Hence my thread to try and get a consensus on what the feature is there for in order to get it restored in a productive way.

[yakra]

Jim's point is that they shouldn't be!

Right, my comment was intended as an acknowledgement of that.

And then we don't get the feature back as we don't come up with a consensus of what it's there for...

Disagree. There's nothing preventing us from implementing an interim solution that, while imperfect, can hold us over while we work on or seek consensus on something better.

[Jim]

Is the dropdown menu of "similar routes" in the showroute on tmtest a sufficient replacement functionality for everyone while we work our if/how to do better on a definition of "related routes"?

[michih]

Is the dropdown menu of "similar routes" in the showroute on tmtest a sufficient replacement functionality for everyone while we work our if/how to do better on a definition of "related routes"?

I think that this minimum addresses my purpose AND oscars + neroute2 needs!
Mine would also apply to stats though...

[oscar]

Is the dropdown menu of "similar routes" in the showroute on tmtest a sufficient replacement functionality for everyone while we work our if/how to do better on a definition of "related routes"?

I think that this minimum addresses my purpose AND oscars + neroute2 needs!
Mine would also apply to stats though...

A stupid question -- where can I find the tmtest version of the HB (and anything else on tmtest)? I'm on the road, haven't had time to look at tmtest.

[Jim]

^^ just start at https://tmtest.teresco.org and navigate around just like the main site.