[OSM-talk] question about untagged green points

[Richard Fairhurst]

Martin Koppenhoefer wrote:

> might be, that I didn't understand exactly how this error is caused,
> but it definitely is a Potlatch-error. Also I can tell you that I was
> quite pleased to see the recent progress of potlatch. It also has some
> cool unique features (like undelete). It is stylish. It is NOT a piece
> of shit. It was not my intention to express this.

Sure, don't worry, I didn't think you had. :)

> But there is some problems that result from not using the API that  
> insert errors in the data.

No. Please, please stop guessing unless you've read the code or talked  
to someone who knows how it works. You keep guessing things which just  
aren't true and confusing people in the process.

As of API 0.6, the AMF API (used by Potlatch) has exactly the same  
constraints as the XML API. There are only two differences.

One is the encoding. XML is textual, AMF is binary. This doesn't have  
any implications for data, only for speed and language support.

The other is the way that the actions are grouped. The XML API, for  
example, groups node, way, and relation fetches for a given bbox in  
the 'map' call. The AMF API splits this out over three calls -  
'whichways', 'getway' and 'getrelation' - which are quicker in the  
live editor environment.

Conversely, the AMF API groups node and way writes in one call,  
'putway', whereas the XML API has them as two separate ones.

All of these, in both APIs, are transactional so a call will either  
succeed in its entirety or not at all.

Now, as it happens, if you have an editor - any editor using any API -  
which saves by making a succession of calls, and you pull the plug out  
from your computer halfway, then yes, some of these calls will have  
been executed and some won't. Conversely, if you have a bug in your  
editor, all the API in the world won't stop wrong (but correctly  
formatted) data being entered. The fact that it speaks XML didn't stop  
JOSM from uploading lat and long the wrong way round!

On the subject of poor connectivity: I know. You might think we all  
have super-fast broadband in the UK, but we don't. I have very poor  
Internet connectivity two-thirds of the time. Flash Player gives the  
client, in this case Potlatch, limited control over the connection in  
such circumstances. As soon as a connection error is reported to  
Potlatch, the little /!\ flashes. It does what it can. Out of  
interest, though, I have never lost data in "edit with save" mode  
because of poor connectivity: I tell it to retry, and it does.

cheers
Richard