Superroutes - good, bad or ugly?

classic Classic list List threaded Threaded
54 messages Options
123
Reply | Threaded
Open this post in threaded view
|

Superroutes - good, bad or ugly?

Paul Allen
I've hesitated to ask this question for months now: what's the
consensus on superroutes?  Going by all I can find on the wiki,
forums and past discussions, they're highly controversial.  One wiki
page mentions them and says don't use them.  They were either
never well documented on the wiki or some documentation has
been scrubbed.  What I don't know is whether the intense dislike
some people expressed when they were first proposed has faded
or if they're still largely considered to be a very bad idea.

One justification for them is that they simplify the mapping of
trans-national routes or very long routes: individual sections can
be mapped separately and then assembled into a coherent whole
as a superroute.

Another justification is that very long ways (the figure I've seen is
300 nodes) can be problematic and they should be split into
a superroute of individual routes.

An argument against them is that some routers may not be able
to handle them.  Which would obviously have been true when they
were first proposed and may still be true now.  Which is a problem
with just about everything proposed here: we propose something
new, it's argued against because editors/renderers/routers don't
handle it, but the reason editors/renderers/routers don't handle it
is because nobody uses it.

The reason I ask is that I can see an application for them that is
not explicitly mentioned in the documentation but might allow me
to deal with an otherwise intractable problem.  If there is universal
disdain here I'll have to abandon that idea.  But if there are enough
people who are happy with them then I have some questions...

Please don't let this degenerate into a flame war.  That can come when
(if) I explain what I want to do with a superroute - even the people who
support superroutes (if there are any) may be unhappy with that idea.

--
Paul


_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
Reply | Threaded
Open this post in threaded view
|

Re: Superroutes - good, bad or ugly?

Andy Townsend
On 13/03/2019 13:18, Paul Allen wrote:
> I've hesitated to ask this question for months now: what's the
> consensus on superroutes?

In what context are you asking the question?  I can think of examples
where the answer would be "a really bad idea" and others where the
answer would be "essential; there's really no other sensible way to have
that data in OSM".

Best Regards,

Andy



_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
Reply | Threaded
Open this post in threaded view
|

Re: Superroutes - good, bad or ugly?

Sergio Manzi
In reply to this post by Paul Allen

If a "superroute" has an official status (like this one: https://www.openstreetmap.org/relation/20773), I'm all-in for that.

If instead it is something "invented" by the mapper, than I'm all-against it.

Can you please provide more information/examples/context?

Sergio


On 2019-03-13 14:18, Paul Allen wrote:
I've hesitated to ask this question for months now: what's the
consensus on superroutes?  Going by all I can find on the wiki,
forums and past discussions, they're highly controversial.  One wiki
page mentions them and says don't use them.  They were either
never well documented on the wiki or some documentation has
been scrubbed.  What I don't know is whether the intense dislike
some people expressed when they were first proposed has faded
or if they're still largely considered to be a very bad idea.

One justification for them is that they simplify the mapping of
trans-national routes or very long routes: individual sections can
be mapped separately and then assembled into a coherent whole
as a superroute.

Another justification is that very long ways (the figure I've seen is
300 nodes) can be problematic and they should be split into
a superroute of individual routes.

An argument against them is that some routers may not be able
to handle them.  Which would obviously have been true when they
were first proposed and may still be true now.  Which is a problem
with just about everything proposed here: we propose something
new, it's argued against because editors/renderers/routers don't
handle it, but the reason editors/renderers/routers don't handle it
is because nobody uses it.

The reason I ask is that I can see an application for them that is
not explicitly mentioned in the documentation but might allow me
to deal with an otherwise intractable problem.  If there is universal
disdain here I'll have to abandon that idea.  But if there are enough
people who are happy with them then I have some questions...

Please don't let this degenerate into a flame war.  That can come when
(if) I explain what I want to do with a superroute - even the people who
support superroutes (if there are any) may be unhappy with that idea.

--
Paul


_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging

_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging

smime.p7s (4K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: Superroutes - good, bad or ugly?

dieterdreist
Am Mi., 13. März 2019 um 14:31 Uhr schrieb Sergio Manzi <[hidden email]>:

If a "superroute" has an official status (like this one: https://www.openstreetmap.org/relation/20773), I'm all-in for that.

If instead it is something "invented" by the mapper, than I'm all-against it.

Can you please provide more information/examples/context?



there are really long relations (e.g. Via Francigena / St. Francis Way, or European E-Routes, which cross half of Europe). Splitting them into smaller pieces helps reducing editing conflicts (2 people editing the same object at the same time). This is often done along administrative boundaries (regional, national).

Cheers,
Martin


_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
Reply | Threaded
Open this post in threaded view
|

Re: Superroutes - good, bad or ugly?

Paul Allen
In reply to this post by Andy Townsend
On Wed, 13 Mar 2019 at 13:29, Andy Townsend <[hidden email]> wrote:
On 13/03/2019 13:18, Paul Allen wrote:
> I've hesitated to ask this question for months now: what's the
> consensus on superroutes?

In what context are you asking the question?  I can think of examples
where the answer would be "a really bad idea" and others where the
answer would be "essential; there's really no other sensible way to have
that data in OSM".

That's more positive than I expected: they're not always on a par with eating babies but the use
has to be justifiable.

Can I get the data into OSM without a superroute?  Sure.  Is that data useful without a superroute?
Not so much.  It is this bus route:
That is incomplete and has some omissions and errors.  I really ought to fix it, but I had
this thought about superroutes and realized if I fixed and then found out I could use a
superroute I'd later have to rework a few things.

It's a circular.  It starts at what can loosely be called the bus station.  It does what can loosely
be called a hairy circular route to return to the bus station.  The route then continues on a side
trip and eventually returns to the bus station, completing the "circle."

There are places where the bus goes into a dead-end and gets out by reversing into a side
junction.  This differs from similar manoeuvres at a terminus of a non-circular route because
passengers are on board.  It does a loop-the-loop.  It appears to do a figure-8 but actually
there are other side-trips that mean it really isn't.

One problem that I don't see a solution for in PTV1, PTV2 or "we don't tag it PTV3" is a stop
that is ignored on the first pass but comes into play on the second pass.  The bus starts at
the bus station A, passes through nodes B, C and D and turns right at D to E.  On this pass
through C it ignores the bus stop there.  After it's gone through the alphabet back to A, it
again goes through B, C and D but this time turns left to alpha, beta, etc.  On this pass it
does stop at C.  Piling all the stops into the relation may lead the routers to conclude that
you can wait at the stop C to get directly to E when you can't (but you can get on at C to take
a detour through the greek alphabet and eventually get to E because it's a circular).

Splitting it into route segments would fix the problem with the stop at C.  On one segment it isn't
a listed stop.  On another segment it is.

Splitting it into route segments would also make it clearer what happens in the loop-the-loop
and the figure-of-8 in the town centre, if the splits are chosen judiciously.  If I'm really clever
I can find splits that make the variant routes fit in nicely, too.  You think that route is insane?
Wait until I add the variants.

Best of all, I could pull these into umap.  It would then be possible to display route segments
in turn to see where the bus goes rather than trying to puzzle it out from the overall route.  Yes,
if you're very familiar with OSM you can puzzle it out from the relation, but most people can't do
that (and I find it difficult, even with knowledge of how the route runs).

So, good idea, bad idea, or should I stick to eating babies as that would be more socially
acceptable?

--
Paul


_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
Reply | Threaded
Open this post in threaded view
|

Re: Superroutes - good, bad or ugly?

Peter Elderson
I can't answer the question for busrelations.

For long hiking routes and walking node networks, relations containing relations are very important.

Without those, maintenance of long hiking routes becomes a p.i.t.b, sometimes near impossible. 

Rendering can be done without superroutes, just by rendering each piece separately. But datausers need to resolve the hierarchy. Waymarkedtrails does that nicely for long recreational routes: Just replace every member relation with its content, recursively. Works nicely- if it's only ways in the lowest level relations. 

I'm not sure if your case needs just rendering or also data use/routing/navigation. 

Vr gr Peter Elderson


Op wo 13 mrt. 2019 om 15:05 schreef Paul Allen <[hidden email]>:
On Wed, 13 Mar 2019 at 13:29, Andy Townsend <[hidden email]> wrote:
On 13/03/2019 13:18, Paul Allen wrote:
> I've hesitated to ask this question for months now: what's the
> consensus on superroutes?

In what context are you asking the question?  I can think of examples
where the answer would be "a really bad idea" and others where the
answer would be "essential; there's really no other sensible way to have
that data in OSM".

That's more positive than I expected: they're not always on a par with eating babies but the use
has to be justifiable.

Can I get the data into OSM without a superroute?  Sure.  Is that data useful without a superroute?
Not so much.  It is this bus route:
That is incomplete and has some omissions and errors.  I really ought to fix it, but I had
this thought about superroutes and realized if I fixed and then found out I could use a
superroute I'd later have to rework a few things.

It's a circular.  It starts at what can loosely be called the bus station.  It does what can loosely
be called a hairy circular route to return to the bus station.  The route then continues on a side
trip and eventually returns to the bus station, completing the "circle."

There are places where the bus goes into a dead-end and gets out by reversing into a side
junction.  This differs from similar manoeuvres at a terminus of a non-circular route because
passengers are on board.  It does a loop-the-loop.  It appears to do a figure-8 but actually
there are other side-trips that mean it really isn't.

One problem that I don't see a solution for in PTV1, PTV2 or "we don't tag it PTV3" is a stop
that is ignored on the first pass but comes into play on the second pass.  The bus starts at
the bus station A, passes through nodes B, C and D and turns right at D to E.  On this pass
through C it ignores the bus stop there.  After it's gone through the alphabet back to A, it
again goes through B, C and D but this time turns left to alpha, beta, etc.  On this pass it
does stop at C.  Piling all the stops into the relation may lead the routers to conclude that
you can wait at the stop C to get directly to E when you can't (but you can get on at C to take
a detour through the greek alphabet and eventually get to E because it's a circular).

Splitting it into route segments would fix the problem with the stop at C.  On one segment it isn't
a listed stop.  On another segment it is.

Splitting it into route segments would also make it clearer what happens in the loop-the-loop
and the figure-of-8 in the town centre, if the splits are chosen judiciously.  If I'm really clever
I can find splits that make the variant routes fit in nicely, too.  You think that route is insane?
Wait until I add the variants.

Best of all, I could pull these into umap.  It would then be possible to display route segments
in turn to see where the bus goes rather than trying to puzzle it out from the overall route.  Yes,
if you're very familiar with OSM you can puzzle it out from the relation, but most people can't do
that (and I find it difficult, even with knowledge of how the route runs).

So, good idea, bad idea, or should I stick to eating babies as that would be more socially
acceptable?

--
Paul

_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging

_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
Reply | Threaded
Open this post in threaded view
|

Re: Superroutes - good, bad or ugly?

Kevin Kenny-3
On Wed, Mar 13, 2019 at 11:03 AM Peter Elderson <[hidden email]> wrote:
> I can't answer the question for busrelations.
(neither can I)

> For long hiking routes and walking node networks, relations containing relations are very important.
> Without those, maintenance of long hiking routes becomes a p.i.t.b, sometimes near impossible.

Exactly. What is 'long' kind of varies. When I mapped
https://www.openstreetmap.org/relation/4286650, I used a single
relation, because there were relatively few member ways and very few
natural points to break it into pieces.

For the work in progress,
https://www.openstreetmap.org/relation/919642, I found that trying to
maintain a single relation was totally unmanageable. In fact, I
tripped over a bug (since fixed) in JOSM with handling a relation that
big, and had to switch to Meerkartor while breaking it up. I decided
that the most natural break was on county lines, although I didn't
trouble with separating off the tiny segment that is on the streets of
Manhattan (New York County). That worked out well, and Waymarked
Trails is happy with it. (The remaining gaps have nothing to do with
tagging. I simply haven't mapped them yet, and won't until the weather
improves. I dislike mapping on snowshoes, because too often I discover
that the snowshoe track doesn't follow the actual treadway.)

> Rendering can be done without superroutes, just by rendering each piece separately. But datausers need to resolve the hierarchy. Waymarkedtrails does that nicely for long recreational routes: Just replace every member relation with its content, recursively. Works nicely- if it's only ways in the lowest level relations.

Most renderers and routers work that way, because that's how osm2pgsql
works. In fact, there is no non-deprecated way to access relation
information at the time of routing or rendering, if you use osm2pgsql;
all relations have been reduced to single ways and/or multipolygons.
(That's why I'm planning to investigate using imposm3 for my own
experimental stuff on route concurrences - I find myself needing the
route relations after osm2pgsql has thrown them away.)

_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
Reply | Threaded
Open this post in threaded view
|

Re: Superroutes - good, bad or ugly?

Graeme Fitzpatrick
In reply to this post by Paul Allen


On Thu, 14 Mar 2019 at 00:05, Paul Allen <[hidden email]> wrote:
or should I stick to eating babies as that would be more socially acceptable?

I guess that sort of depends on whether or not they're Irish babies https://en.wikipedia.org/wiki/A_Modest_Proposal

& we're having Roast Peasant under Glass for dinner tonight - feel like joining us? :-)

& as for super-relations - ??? :-) 

Thanks

Graeme

_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
Reply | Threaded
Open this post in threaded view
|

Re: Superroutes - good, bad or ugly?

Jo-2
I think we should move to subrelations for bus routes at some point. Actually doing it is somewhat tricky. We'd definitely need editor support to show that a route which consists of subroutes is continuous or not. The biggest point of contention seems to be whether the stops should go into the subroute relations as well. I'd say no. Keep the stop sequences for a particular variation in itinerary together in a route relation per variation. And only use the subroutes to collect continuous sequences of ways.

At the same time, some people would think it's more logical to have a sequence of ways, then a stop, then the next sequence, a stop, and so on. For that too, it would be nice to have editor support that shows that the way before and the way after actually connects, or not.

If we go the way of subroutes for PT routes, I/we can probably coerce a GSoC student to create support for it in PT_Assistant over the summer.

Polyglot

On Wed, Mar 13, 2019 at 11:20 PM Graeme Fitzpatrick <[hidden email]> wrote:


On Thu, 14 Mar 2019 at 00:05, Paul Allen <[hidden email]> wrote:
or should I stick to eating babies as that would be more socially acceptable?

I guess that sort of depends on whether or not they're Irish babies https://en.wikipedia.org/wiki/A_Modest_Proposal

& we're having Roast Peasant under Glass for dinner tonight - feel like joining us? :-)

& as for super-relations - ??? :-) 

Thanks

Graeme
_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging

_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
Reply | Threaded
Open this post in threaded view
|

Re: Superroutes - good, bad or ugly?

Warin
In reply to this post by dieterdreist
On 14/03/19 00:36, Martin Koppenhoefer wrote:
Am Mi., 13. März 2019 um 14:31 Uhr schrieb Sergio Manzi <[hidden email]>:

If a "superroute" has an official status (like this one: https://www.openstreetmap.org/relation/20773), I'm all-in for that.

If instead it is something "invented" by the mapper, than I'm all-against it.

Can you please provide more information/examples/context?



there are really long relations (e.g. Via Francigena / St. Francis Way, or European E-Routes, which cross half of Europe). Splitting them into smaller pieces helps reducing editing conflicts (2 people editing the same object at the same time). This is often done along administrative boundaries (regional, national).


 Supper relation 176684 the Bicentennial National Trail is over 5,000 km long, split into 4 sub relations at present.


Many bus routes in cities follow the same route in small sections. It would be hand for them to share that part as a sub route - meaning any updating of that sub route (people adding turn restrictions, parking, curbing etc) only changes that sub route rather than 4 or more individual bus routes.


--------------

The splitting of a relation into sub relations is done for many reasons. Making it easier for the mapper is of primary importance. I did see on the wiki that relations are better with a maximum of 300 members .. weather that is true or not I don't know. But it doe make sense that with a lot of members the relation becomes harder to handle for both maintenance and rendering.


_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
Reply | Threaded
Open this post in threaded view
|

Re: Superroutes - good, bad or ugly?

Paul Allen
In reply to this post by Jo-2
On Wed, 13 Mar 2019 at 22:42, Jo <[hidden email]> wrote:
I think we should move to subrelations for bus routes at some point. Actually doing it is somewhat tricky. We'd definitely need editor support to show that a route which consists of subroutes is continuous or not.

Not a big problem.  Not compared with sorting the route I just had to sort (again).  Dunno if I messed
up last time I sorted it or what, but it's a major pain.  Maybe I need a different JOSM plugin, but I
already find JOSM confusing enough.

One problem I found with JOSM was zooming to a way in the relation when that way appears more
than once.  I did that a lot so I could check the next way in the route was even in the table or
had to be added.  But if a way appears more than once in a route (as several do in the one
I was working on) then JOSM moves to the first occurrence of it in the table of ways, rather than
the one I just selected.

If I could use a superroute then I would choose the subroutes so that no way appears twice
in a subroute.  That's a problem with circulars that non-circulars don't have (most of the time
a non-circular has two variants A->B and B->A and although most of the ways are common to
both they're separate relations).  A truly circular wouldn't have that problem either, but this one
has loops and repeats and side-branches that aren't loops.
 
The biggest point of contention seems to be whether the stops should go into the subroute relations as well. I'd say no. Keep the stop sequences for a particular variation in itinerary together in a route relation per variation. And only use the subroutes to collect continuous sequences of ways.

At the same time, some people would think it's more logical to have a sequence of ways, then a stop, then the next sequence, a stop, and so on. For that too, it would be nice to have editor support that shows that the way before and the way after actually connects, or not.

That's the other problem I have.  A stop which isn't a stop the first time the bus passes along
a way but is a stop the second time.  Although the ordering of stops does mean it's possible to
figure out if you inspect the relation closely, it's not obvious.  Allowing stops to intersperse the
ways would make it slightly easier to figure out.  But using a superroute makes it very clear:
it's not a stop on segment 1 but it is a stop on segment 2.  So I'd definitely want stops on the
subroutes not the superroute.  I think it makes more sense that way anyway.  It keeps the
stops with the associated ways.

If we go the way of subroutes for PT routes, I/we can probably coerce a GSoC student to create support for it in PT_Assistant over the summer.

All the tools I need to make life a LOT easier by splitting the route into subroutes are there.
Breaking the route up manually is relatively simple (takes several steps and some time, but
very little thought), and after that it just gets far, far easier to deal with short segments than
one big, looping, wandering route.  If I choose the subroutes wisely I can even let JOSM sort
them for me (it gets very confused with this route where several ways are traversed more than
once).

There are only three things stopping me doing that right now:

1) What the response to my post was.  So far nobody has said that it's a really bad idea since
I posted details of the route.  Unless there are many howls of outrage, I'll assume it's not
an outrageously bad idea.

2) How to differentiate between the different segments.  I can't use from and to because those are
supposed to be terminal destinations (preferably as shown on the bus itself).  So I need a new
tag (or two).  I could get by with something like segment="A to B", segment="B to C" etc.  But
there's probably a better way of doing it.  And I'd like to do that both for human readability when
mapping and to allow umap a way of distinguishing between subroutes with an overpass
query.

3) What else I haven't thought of.  What's show-stopper might turn up after I've spent time
splitting the route.

--
Paul


_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
Reply | Threaded
Open this post in threaded view
|

Re: Superroutes - good, bad or ugly?

Paul Johnson-3
In reply to this post by Paul Allen
On Wed, Mar 13, 2019 at 8:19 AM Paul Allen <[hidden email]> wrote:
I've hesitated to ask this question for months now: what's the
consensus on superroutes?  

Coherently and cogently mapping large countries with long routes (such as the United States) would be essentially impossible without them.  I think the current relation member limit is arbitrarily set to what fits in an 8 bit value IIRC; throw in everything that can cause a break (bridges, changes in weight, speed and length maximums and minimums, whether or not something is toll, lane counts and placement, turn lane status, etc) and it wouldn't surprise me if states like Texas or California might have to start using county-by-county child relations within a superrelation in order to keep it all under the member limit in the foreseeable future.

_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
Reply | Threaded
Open this post in threaded view
|

Re: Superroutes - good, bad or ugly?

Warin
In reply to this post by Paul Allen
On 14/03/19 01:02, Paul Allen wrote:


One problem that I don't see a solution for in PTV1, PTV2 or "we don't tag it PTV3" is a stop
that is ignored on the first pass but comes into play on the second pass.  The bus starts at
the bus station A, passes through nodes B, C and D and turns right at D to E.  On this pass
through C it ignores the bus stop there.  After it's gone through the alphabet back to A, it
again goes through B, C and D but this time turns left to alpha, beta, etc.  On this pass it
does stop at C.  Piling all the stops into the relation may lead the routers to conclude that
you can wait at the stop C to get directly to E when you can't (but you can get on at C to take
a detour through the greek alphabet and eventually get to E because it's a circular).
IN PTV2 you list the stops in order .. so they would be listed as;
A
B
D
E
etc
A
B
C
D
E
etc

So it can be done in PTV2.

Splitting it into route segments would fix the problem with the stop at C.  On one segment it isn't
a listed stop.  On another segment it is.

A segment end does not indicate a stop .. in PTV2. The segments need to be in sequential order and so do the stops.

I did a rough simpletons guide to PTV2 .. https://www.openstreetmap.org/user/Warin61/diary/45106



_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
Reply | Threaded
Open this post in threaded view
|

Re: Superroutes - good, bad or ugly?

TonyS

Warin

Great description of PTV2.

Paul

Am I right in thinking that a superroute is a sequence of ways and relations of ways? The relation of ways could be called a route-segment or similar. A I see it routes for most trains and buses are a sequence of ways and route-segments, and a route-segment could be used by many routes.

TonyS

On 14/03/2019 03:43, Warin wrote:
On 14/03/19 01:02, Paul Allen wrote:


One problem that I don't see a solution for in PTV1, PTV2 or "we don't tag it PTV3" is a stop
that is ignored on the first pass but comes into play on the second pass.  The bus starts at
the bus station A, passes through nodes B, C and D and turns right at D to E.  On this pass
through C it ignores the bus stop there.  After it's gone through the alphabet back to A, it
again goes through B, C and D but this time turns left to alpha, beta, etc.  On this pass it
does stop at C.  Piling all the stops into the relation may lead the routers to conclude that
you can wait at the stop C to get directly to E when you can't (but you can get on at C to take
a detour through the greek alphabet and eventually get to E because it's a circular).
IN PTV2 you list the stops in order .. so they would be listed as;
A
B
D
E
etc
A
B
C
D
E
etc

So it can be done in PTV2.

Splitting it into route segments would fix the problem with the stop at C.  On one segment it isn't
a listed stop.  On another segment it is.

A segment end does not indicate a stop .. in PTV2. The segments need to be in sequential order and so do the stops.

I did a rough simpletons guide to PTV2 .. https://www.openstreetmap.org/user/Warin61/diary/45106



_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging

_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
Reply | Threaded
Open this post in threaded view
|

Re: Superroutes - good, bad or ugly?

Paul Allen
In reply to this post by Warin
On Thu, 14 Mar 2019 at 03:44, Warin <[hidden email]> wrote:
On 14/03/19 01:02, Paul Allen wrote:


One problem that I don't see a solution for in PTV1, PTV2 or "we don't tag it PTV3" is a stop
that is ignored on the first pass but comes into play on the second pass.  The bus starts at
the bus station A, passes through nodes B, C and D and turns right at D to E.  On this pass
through C it ignores the bus stop there.  After it's gone through the alphabet back to A, it
again goes through B, C and D but this time turns left to alpha, beta, etc.  On this pass it
does stop at C.  Piling all the stops into the relation may lead the routers to conclude that
you can wait at the stop C to get directly to E when you can't (but you can get on at C to take
a detour through the greek alphabet and eventually get to E because it's a circular).
IN PTV2 you list the stops in order .. so they would be listed as;
A
B
D
E
etc
A
B
C
D
E
etc

So it can be done in PTV2.

Yes, it can be done.  I said as much myself.  But it is hard to decipher.  There is nothing that
explicitly says that the bus does not  stop at C the first time the bus passes it but does stop
the second time.  If you painfully trace out the ENTIRE route, correlating stops and ways,
you can reach the correct conclusion.

Now consider somebody using the query tool for the bus stop at C.  Somebody who isn't a
mapper, just a user.  Here's the route:
and here's the stop: https://www.openstreetmap.org/node/5706768172 Note that I haven't yet
included the stop on that route (although I have included it in an incomplete variant route), but
if I had, and included it in the right order, how easy would it be for you to figure out what is going on?
From casual inspection, it's not easy.

However, take a look at the incomplete variant route: https://www.openstreetmap.org/relation/8603360#map=14/52.0869/-4.6691
and imagine that were part of a superroute.  You could see, by inspecting the subroutes in turn
(for example, on umap) that this stop is used on one route segment but not another, even though
both segments pass along that way.

A segment end does not indicate a stop .. in PTV2. The segments need to be in sequential order and so do the stops.

A segment END?  Who's talking about segment ends?  I'm talking about a segment which may
have more than one stop or no stops at all.  Who's talking about disordered segments?  The whole
idea is to have the segments in order.  Who's talking about disordered stops?  The whole idea is
to have ordered stops.

What I'm talking about is a segment of a route with its stops.  It may share one or more ways
with a different segments of that route, and it may share one or more stops (or none at all).  One
of us is missing what the other is saying, right now I don't know which one of us that is.

--
Paul


_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
Reply | Threaded
Open this post in threaded view
|

Re: Superroutes - good, bad or ugly?

Paul Allen
In reply to this post by TonyS
On Thu, 14 Mar 2019 at 11:21, Tony Shield <[hidden email]> wrote:

Am I right in thinking that a superroute is a sequence of ways and relations of ways?

I'm not 100% certain.  The documentation on the wiki isn't entirely clear.  I suspect some of it
may have been scrubbed by those who dislike the concept of superroutes for any purpose
whatsoever, but it may just never have been written in the first place.

As I understand it (possibly entirely inccorectly) it would be analogous (using a slightly
different mechanism) to having the Wales Coast Path be a relation comprising the
route for the Pembrokeshire Coast Path, the Ceredigion Coast Path, etc.  A relation of
relations.
 
The relation of ways could be called a route-segment or similar. A I see it routes for most trains and buses are a sequence of ways and route-segments, and a route-segment could be used by many routes.

I wasn't intending to share segments between routes.  I'm not sure if I could make that work for
some of my purposes (I haven't thought about it enough to be sure).  It probably won't be
universally applicable anyway, there are going to be routes that share a particular segment but
not all of the stops on that segment.  Also fan-out of routes is going to result in a lot of very
tiny segments that are possibly more trouble than they're worth and it's probably less work to
make them longer segments.

--
Paul


_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
Reply | Threaded
Open this post in threaded view
|

Re: Superroutes - good, bad or ugly?

Jo-2
I would definitely want routes to be composed of subroutes which are shared with other routes, hence the reasoning of keeping the stop sequences in the route relations.

Polyglot

On Thu, Mar 14, 2019, 15:41 Paul Allen <[hidden email] wrote:
On Thu, 14 Mar 2019 at 11:21, Tony Shield <[hidden email]> wrote:

Am I right in thinking that a superroute is a sequence of ways and relations of ways?

I'm not 100% certain.  The documentation on the wiki isn't entirely clear.  I suspect some of it
may have been scrubbed by those who dislike the concept of superroutes for any purpose
whatsoever, but it may just never have been written in the first place.

As I understand it (possibly entirely inccorectly) it would be analogous (using a slightly
different mechanism) to having the Wales Coast Path be a relation comprising the
route for the Pembrokeshire Coast Path, the Ceredigion Coast Path, etc.  A relation of
relations.
 
The relation of ways could be called a route-segment or similar. A I see it routes for most trains and buses are a sequence of ways and route-segments, and a route-segment could be used by many routes.

I wasn't intending to share segments between routes.  I'm not sure if I could make that work for
some of my purposes (I haven't thought about it enough to be sure).  It probably won't be
universally applicable anyway, there are going to be routes that share a particular segment but
not all of the stops on that segment.  Also fan-out of routes is going to result in a lot of very
tiny segments that are possibly more trouble than they're worth and it's probably less work to
make them longer segments.

--
Paul

_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging

_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
Reply | Threaded
Open this post in threaded view
|

Re: Superroutes - good, bad or ugly?

TonyS

Hi

Is there a PTV2 example of route with ways and subroutes? Can they currently be edited with JOSM?

TonyS

On 14/03/2019 15:07, Jo wrote:
I would definitely want routes to be composed of subroutes which are shared with other routes, hence the reasoning of keeping the stop sequences in the route relations.

Polyglot

On Thu, Mar 14, 2019, 15:41 Paul Allen <[hidden email] wrote:
On Thu, 14 Mar 2019 at 11:21, Tony Shield <[hidden email]> wrote:

Am I right in thinking that a superroute is a sequence of ways and relations of ways?

I'm not 100% certain.  The documentation on the wiki isn't entirely clear.  I suspect some of it
may have been scrubbed by those who dislike the concept of superroutes for any purpose
whatsoever, but it may just never have been written in the first place.

As I understand it (possibly entirely inccorectly) it would be analogous (using a slightly
different mechanism) to having the Wales Coast Path be a relation comprising the
route for the Pembrokeshire Coast Path, the Ceredigion Coast Path, etc.  A relation of
relations.
 
The relation of ways could be called a route-segment or similar. A I see it routes for most trains and buses are a sequence of ways and route-segments, and a route-segment could be used by many routes.

I wasn't intending to share segments between routes.  I'm not sure if I could make that work for
some of my purposes (I haven't thought about it enough to be sure).  It probably won't be
universally applicable anyway, there are going to be routes that share a particular segment but
not all of the stops on that segment.  Also fan-out of routes is going to result in a lot of very
tiny segments that are possibly more trouble than they're worth and it's probably less work to
make them longer segments.

--
Paul

_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging

_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging

_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
Reply | Threaded
Open this post in threaded view
|

Re: Superroutes - good, bad or ugly?

dieterdreist


sent from a phone

> On 14. Mar 2019, at 16:49, Tony Shield <[hidden email]> wrote:
>
> Can they currently be edited with JOSM?


of course, you simply add a relation as member to another relation.

Cheers, Martin

_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
Reply | Threaded
Open this post in threaded view
|

Re: Superroutes - good, bad or ugly?

Paul Allen
In reply to this post by Jo-2
On Thu, 14 Mar 2019 at 15:09, Jo <[hidden email]> wrote:
I would definitely want routes to be composed of subroutes which are shared with other routes,

I see that as less than useful for any route I know of.  But I suppose it's a matter of how short
a subroute you're willing to put up with.  I probably wouldn't do it myself but would say that you
should not.  However, surely that means that the subroutes also have to be anonymous (they
don't specify a ref or anything else specific to a particular route) and would therefore be totally
unsuitable for me.  If that's the case then I'm not in favour of doing things your way.
 
hence the reasoning of keeping the stop sequences in the route relations.

That, as I understand it, goes against what is currently done in superroutes, goes totally
against one of my main reasons for wanting to do this, and makes editing far harder.  Keeping
the stops with the route segment makes sense for many reasons.  It means subroutes can
be edited independently of one another.  It makes adding/removing stops easier for the simple
reason that it's easier to figure out where a stop goes in a relatively short subroute than on
a much longer superroute.  It means the standard carto query on a subroute shows the stops
associated with that subroute whereas your way you'd have to query the superroute to see
the entire route and all its stops in order to see the stops on a much smaller subroute.

All in all, I think keeping stops in the superroute is a very bad way to go, independent of it
achieving precisely the opposite of what I hope to do.

Your "enhancement" makes the whole idea useless for my purposes and, I think, is impracticl
for any other purpose.

--
Paul


_______________________________________________
Tagging mailing list
[hidden email]
https://lists.openstreetmap.org/listinfo/tagging
123