[Geojson] Semantics of adjacency in GeoJSON text sequences

[Sean Gillies <[e-mail]>]

Dear all,

In https://github.com/geojson/geojson-text-sequences/issues/13 Martin Daly
asks what to do about extra members of a GeoJSON FeatureCollection when
converting application/geo+json to application/geo+json-seq (GeoJSON text
sequence) and back. I don't know what the best practice is but hope that
we, the WG, can work that out.

What I'd like to discuss here is the semantics of adjacency in sequences in
the context of round-tripping GeoJSON <-> GeoJSON Text Sequences (let's
just say "sequence"). Do we mean anything when we write a FeatureCollection
object to a sequence and then write a Feature object? Do we mean the
Feature is part of the FeatureCollection? If we then write another
FeatureCollection to the sequence, do following Features "belong" to that
FeatureCollection, or to the first, or to neither?

In the draft right now it says that a GeoJSON text sequence can contain any
kind of GeoJSON object. It also says that sequences with significant order
are less interoperable than those with no significant order. This is
because in a distributed system a sequence might have any number of writers
and the order of records will be difficult to guarantee. This uncertainty
in order makes it hard for sequence adjacency to mean anything. I'm
concerned that anything we do to make adjacency meaningful will make
sequences less reliable in distributed systems than they could be.

I'm leaning toward strengthening the interop guidance and adding something
like "adjacency has no meaning and implies no relationship" to the
definition of GeoJSON text sequences. Is anyone opposed to that or see
something that I'm missing?

--
Sean Gillies