Closed derhuerst closed 7 years ago
@juliuste argued that there's actual value in clusters being able to overlap each other, which indicates a m:n
relationship with station
s.
As @derhuerst already mentioned, I'd strongly advocate a m:n
relationship in order to support - or at least to not disable by default - the possibility of assigning multiple cluster
s to one station
. For instance: This could be useful when clustering station
s into both geographical regions and cities. The station Nice-Ville
could then be included in the clusters named Cote d'Azur
and Nice
which - at least to me - really makes sense.
Regarding the naming issue, I'd personally choose between cluster
and region
. cluster
is the more general term while region
would probably be more easily understandable for people who didn't read the specification when using one of our modules - at least I'd know what do do with a regions
method while I might not be so sure what a method called cluster
does. This isn't the strongest or most important argument, still not completely irrelevant though.
Anyway, I wouldn't call the new type hub
, since this sounds more like an important interchange station than an actual cluster. I'd even call big stations like Frankfurt Hbf
or - even bigger - Frankfurt Flughafen
a hub on their own since they serve as an interchange point for quite a decent amount of connections.
On where to specify the relations: As @derhuerst already mentioned in a private discussion, there is no way to reverse engineer the station-cluster-relation starting at a cluster in a lot of cases. Therefore, I'd propose to at least include a regions
key in station
object. For practical reasons though, I'd also like to have an additional optional stations
key in region
objects to enhance the workflow in some situations.
Bump 🔨
Therefore, I'd propose to at least include a
regions
key instation
object. For practical reasons though, I'd also like to have an additional optionalstations
key inregion
objects to enhance the workflow in some situations.
Sounds reasonable to me. However, we should clearly stress what's the preferred direction if only one direction can be specified.
When @juliuste migrated
meinfernbus
to FPTF, we noticed that there is the concept of regions/cities/hubs.Berlin for example has several long-distance train & bus stations, all distinct but well-connected through local public transport. I makes sense to keep them as stations, because they may still have individual stops, but clustering them would enable more advanced routing information.
Now questions arise:
1:n
orm:n
?stations
field or the opposite? or both?