[Tilesathome] Questions on tile updating process

[Andre Hinrichs]

On Montag 20 April 2009, Dirk-Lüder Kreie wrote:
> Andre Hinrichs schrieb:
> > Here are the checks it does in order:
> > 1.) Tile has an unhandled request -> ignore
> > 2.) Tile has an active request -> ignore
> > 3.) Tile is empty and was recently finished -> ignore

Hmm... with the short format request I cannot even see, if an empty tile was 
recently finished. Is there a short request to get the render requests of a 
tile? Only the status is needed: unhandled, active, finished...

> > 4.) Tile is empty and sea -> ignore
>
> OK so far
>
> > 5.) Tile is empty and land -> rerender
>
> Why? If there was an edit it would have hit the autorequester...

Because of two reasons.
Firstly, there are often long ways (ferry,border,roads) and the nodes lay 
outside of the tile.
Secondly there are currently huge areas totally blank because of the TRAPI 
error a few weeks ago. First I went through the map optically which was 
absolutely nuts especially because afterwards the server crash came and all 
work was lost.

Well, I see that many empty land tiles are really empty and was thinking to 
only rerender a tile if a neighbour tile has content. But this would also make 
it necessary to get the render status of a tile a while after a request has 
been sent to make sure a request has finished.

> > 6.) Tile is empty and mixed -> rerender
>
> likewise

Same reason.

> > 7.) Tile is empty and unknown -> ignore with warning
> > 8.) Tile is no older than a given period -> ignore
>
> OK
>
> > 9.) Tile is older than a given period -> rerender
>
> 6-12 months, see below.

After the initial phase of the database this would go much faster.
Also I think that I can ignore all sea tiles completely during initial phase. 
I would prefer a walk-along-the-line procedure in a later phase for borders 
and ferry routes.

> > 10.) Tile is not empty and has no date -> ignore with warning
>
> OK
>
> If you go through every one of the 16777216 tiles at a rate of one per 2
> seconds (as not to interfere too much with the normal operation of the
> server) you will take more than a year, more than 6 months if you do it
> at a more aggressive rate of one request per second.

Well, I was thinking of sleeping after each request exactly the duration the 
request has taken. This would automatically reduce load from the script if the 
server is very busy. The longer the request takes the longer the script 
sleeps.


Regards,
Andre