[OSM-dev] Removing functionality and giving just No as answer

Tom Hughes tom at compton.nu
Wed Mar 8 12:03:32 UTC 2017

On 08/03/17 09:35, joost schouppe wrote:

> While a lot of the comments are a bit misguided, I think it is clear
> that quite a few mappers use this dirty trick to get the render to
> refresh. As OSM.org is supposed to be a mapper's tool and not a general
> public website, I think it is quite obvious that a forced-tile-refresh
> function is within the scope of the website.

There is a cargo cult belief that you need to force dirty the tile at 
the renderer but that is almost never the case - any changes to nodes 
and ways will automatically dirty the relevant tiles. Only changes to 
relations are not handled automatically.

Now there was a bug for much of last week where that wasn't happening 
properly - that was fixed on Friday morning and all the missed tile 
expiries were processed over the next 24 hours.

What is normally an issue is the squid caches that sit in front of the 
renderers, but a simple shift-reload will bust those caches and in fact 
a manual dirty will do nothing at all to help with those.


Tom Hughes (tom at compton.nu)

More information about the dev mailing list