My best guess is there's a problem in serving the map data up to users.
Vector slippy maps run on bounding box searches - if I look at New York, my client says to the server "Tell me all the roads within this square" with the latitude and longitude of wherever I'm looking.
There's a lot of data - too much to store on one server - and the data doesn't all shard naturally. You can divide the world up into squares and store different squares on different servers, but long roads cross many squares, big objects like country borders belong in many squares, and when you're zoomed out your view covers many squares.
The server also has to be clever; I don't actually want every road in New York or the data would take forever to load. I only want the major roads when I'm zoomed out.
All these problems are solvable - spatial databases like PostGIS have been able to perform bounding box queries for years - but I don't know of anyone using PostGIS to support the millions of users Google have. It's well known that for regular data Google has things like BigTable, fancy NoSQL stuff that's supposed to scale well; I assume they have a geospatial database along similar lines. It's possible Apple decided to develop something similar.
My guess as to how [1] came about is that the user started zoomed out, so they only got sent major roads, then when they zoomed in some hiccup prevented loading the minor roads. This could have been as simple as the user's wifi going down, or it could that Apple's database servers are overloaded or have bugs.
Vector slippy maps run on bounding box searches - if I look at New York, my client says to the server "Tell me all the roads within this square" with the latitude and longitude of wherever I'm looking.
There's a lot of data - too much to store on one server - and the data doesn't all shard naturally. You can divide the world up into squares and store different squares on different servers, but long roads cross many squares, big objects like country borders belong in many squares, and when you're zoomed out your view covers many squares.
The server also has to be clever; I don't actually want every road in New York or the data would take forever to load. I only want the major roads when I'm zoomed out.
All these problems are solvable - spatial databases like PostGIS have been able to perform bounding box queries for years - but I don't know of anyone using PostGIS to support the millions of users Google have. It's well known that for regular data Google has things like BigTable, fancy NoSQL stuff that's supposed to scale well; I assume they have a geospatial database along similar lines. It's possible Apple decided to develop something similar.
My guess as to how [1] came about is that the user started zoomed out, so they only got sent major roads, then when they zoomed in some hiccup prevented loading the minor roads. This could have been as simple as the user's wifi going down, or it could that Apple's database servers are overloaded or have bugs.
[1] http://news.bbcimg.co.uk/media/images/63010000/jpg/_63010615...