[OSM-talk] Best base to build on ...

Lester Caine lester at lsces.co.uk
Sun Aug 16 10:12:36 UTC 2015

On 16/08/15 01:46, Paul Norman wrote:
> On 8/15/2015 2:09 PM, Lester Caine wrote:
>> The simple answer seems to be that there is no standard when it comes to
>> mapping applications and everybody creates their own personal special
>> such as kosmtik rather than working with established standards:(
> Kosmtik is a standard. But it's a development tool for stylesheet
> design. The other options are Mapbox Studio and Tilemil. All three are
> essentially desktop applications, although the developer using them
> might interact with them via a browser.
> The former doesn't support a raster tile style, has tie-ins with Mapbox
> that make it difficult to use independently, and does not work
> reasonably with styles in version control. The problem with Tilemill is
> that is is abandoned, and includes in-program text editing, which adds
> significant complexity to the codebase, and this text editing does not
> function with large complex styles.
> It's also not a question of duplication, as all of these options are
> written in nodejs and the map rendering parts share common modules.
> renderd, and all similar software, serves a completely different
> purpose, and are not good options for applications where Kosmtik,
> Tilemill, and Mapbox Studio make sense.

I'm slowly getting things back under control and realising that some
parts of the jigsaw just need to be put into different boxes. And some
of the boxes I'm happier with are not perhaps the best options for
working with OSM.

Being a long time PHP developer and C/C++ for many years before that,
'Ruby/Rails' is just off the list. Python came on board because of tools
like mercurial, and I do even dig into java to maintain the eclipse
based environment. So node is yet another box which is on the edge of
things and javascript is an essential tool with PHP.

The switch to vector tiles was suggested on the GB list, then sort of
pulled as 'not being ready for general use', but my short term goal is a
source of tiles in the current style ... so raster ones ... before they
get lost from OSM.

Life would probably be easier if I rolled back from nginx to apache, but
the performance difference on services that I get paid for make that a
no-go, so I need to be rendering tiles nginx can cache and I've sort of
cracked that in a manor that seems to be working although updating the
cache may be a little slow. But I though 1000 days to create a full set
may b a little slow :)

Steve Bennett seems to have cracked all this, but WTF is 'Salt'. His
scripts have at least pointed in the right direction and I even have the
older osrm instance merged so I can view things.

What I don't seem to want to be doing is going for options that rely on
mbtiles? Or am I missing something, and an mbtile database is more
efficient than a directory structure of tiles where an area the size of
the UK is involved?

Lester Caine - G8HFL
Contact - http://lsces.co.uk/wiki/?page=contact
L.S.Caine Electronic Services - http://lsces.co.uk
EnquirySolve - http://enquirysolve.com/
Model Engineers Digital Workshop - http://medw.co.uk
Rainbow Digital Media - http://rainbowdigitalmedia.co.uk

More information about the talk mailing list