Notices

Well its been a year and a half since the last post. I wonder if anyone has done anything with roadnav. Roadmap is no where near as good. I hope this gets jumped on soon.

Now that Android has been ported to Nit devices and also Sygic is releasing Mobile Maps for Android i think that maybe (as long as GPS is supported in the Android port) the GPS navigation software issue is all going to be taken of.

Okay, probably wrong forum, old post, etc., but it got me to thinking...

Why can't Maemo Mapper generate tiles from local vector data on the fly? I mean, basic vector data for a whole lotta' places is in the Mb range, not Gb. Ideally, mapper would have a configurable switch that just says "generate missing tiles." When you get to an area/zoom where tiles aren't cached, and you can't download the tile from a repository (offline), it just makes the tile on the fly from the local vector data. It can't be THAT hard. Even better (feature creep here) that vector data could also be used for locally-calculated routes.

On a more basic level, how about having a "local repository" in Mapper that is just a separate program that generates tiles on the fly. If you want tile X, generate it from whatever vector data is cached. OSM data is getting to be good enough for this now, so licensing costs are nil.

Why not? It would be the best of both...

It's not that easy, it's maps we're talking about. They have reference systems, projections (damn that round Earth), datums, etc, which make rasterization a wee bit more complicated than just making a vector painter. The 'real' software that does that is Mapserver or Mapnik, but that's heavy artillery.

ya-but... I know it's not trivial, and I know about datums, spheroids, and the like, (well, never done the math or the code, but I've worked enough with various mapping apps, doing the conversions). But, if you consider that Navit does it, Roadmap does it, even my little Garmin receiver with a whopping 8Mb or RAM does it, so the problem is far from intractable. It also doesn't have to be done nearly as fast as something that's relying on vector data real-time - like everything that uses vector data.

What I'm trying to say is that if my literally-stupid little Garmin receiver can take compiled vector data and then generate a raster display in less time than it takes for me to get bored, scream at it, and throw it out the window... well, it can't take that much processing power.

As there is already an open-source code base (Navit et-all) that processes vector data, and reasonably good OSM map data, all that really needs to have happen is:

* a local process that receives "requests" for tiles and then generates them from local vector data.

* Mapper to have a special repository that's not a webpage, but this local process.

I know it's not "easy" (I'd do it myself if it was - I like easy). But, it's not like a whole mapping program would have to be created, or Hildonized... it's a background process that would start alongside Mapper that just responded to "web-page-like" requests. It would eliminate one of the bigger complaints about Mapper (no cached maps and not online = nothing) while leaping past the existing vector-map offerings with all the great benefits that Mapper has to offer. It also offers a path (pun intended) towards offering local routing. And, yes, I know routing is a much harder problem - though again, hand-held Garmins do it - so it's possible.

Just thinking out loud is all...

Just a follow up...

It could just be a stand-alone program that takes a location and path as input and then generates tiles right into the Mapper database and cache while Mapper is off. That would be the simplest starting point and wouldn't even need any mods to Mapper itself. Sort of like the Map Tile Downloader but, rather than downloading, it just generates them from local vector data.

Hmmm, set mapper to "repository Local" that has a dummy URL. Then, run program to fill said repository with data, at least around a specified location and to a specified zoom. Somewhat of a kludge, but a good first step. I wonder what the Navit code-base looks like...

Still thinking out loud...