Sam Vekemans wrote:

> Bonkers? ... perhaps :@)

Well if I've understood correctly then what you're suggesting is that we 
add what amounts to presumably some tens of millions of nodes each of 
which would presumably have a large number of highly duplicative tags.

If that is what you're saying then I stand by my comment that it seems 
completely and utterly mad.

> Actually, that would be the just the 1st step. meaning that;  On those 
> mapped areas where say - if the actual shapes were imported, you would 
> see 'ghost lines'. .. this is not preferred.  So, by just seeing the 
> reference points, then these shapes can be drawn in by hand (where needed).

If everything is going to be drawn by hand then why bother importing 
stuff at all? Why not just create a tool that can generate an OSM file 
for an area from Geobase and then let people generate extracts for their 
area and load them into JOSM to do the hand addition of the lines before 
uploading?

> Then, on the areas where it has been just OSM roads done, the rest of 
> the import can happen, so the connect-the-dots would just be for the roads.

Ah right, so you're just suggesting this for everything, just for some 
subsets of the data in each area? How will you decide which features to 
import fully and which to import partially in each area?

> Going this way, it would avoid the number 1 objection to the import, 
> which is importing data showing up as ghost lines, because the 
> change-detect script wasn't able to detect that someone already drew in 
> the map feature.

What are ghost lines? What is this change-detect script and what is it 
going to be doing?

> and going this way, would also solve the problem of how to deal with 
> updates..... they would get imported as nodes instead of ways.  (so not 
> to interfere with any mappers work)

To be honest I think you're probably living in a dream world if you 
think there is any real chance of being apply to apply updates from the 
base data automatically.

> So far, it looks like the majority of the country would be able to 
> accept a full import of everything... except the roads. .. it's only 10 
> or so tiles that nothing would be imported (just nodes), and less than 
> 100 where only roads would be omitted.  (there are 12,539 tiles in 
> Canada) (.5x.25 degrees) - something like that... the actual breakdown 
> will (should) be posted on the wiki.

So this is mostly only about the roads, but in a lot of ways people 
would probably consider them the most important feature, so if we don't 
get any roads imported without them having to be traced over then we've 
only gained a fairly limited amount.

To be honest most of this is obviously up to you guys as you're the ones 
that will have to work with it. My main concern is what impact whatever 
scheme you choose will have on the server infrastructure and data 
storage requirements, which is why I have concerns about creating tens 
of millions of nodes which have many duplicated tags or which will wind 
up never being used and will just be deleted again.

Tom

-- 
Tom Hughes (t...@compton.nu)
http://www.compton.nu/

_______________________________________________
Talk-ca mailing list
Talk-ca@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-ca

Reply via email to