[OSM-dev] Verify upload of trackpoints/segments

[bvh]

So I've added uploading trackpoints and segments to Openstreetmap
in Merkaartor.  Could someone verify that I didn't make any gross
mistakes? You can find some sample work I did
around 51.9157 Lat 4.43408 Lon. It's the triangular structure.

I've made available a new version of Merkaartor and now also provide
a Windows binary for those who do not have a Qt4 development
environment available. All at http://www.irule.be/bvh/c++/merkaartor
Feel free to dis/encourage me some more.

Now I have a few more questions:

1. tags

At http://wiki.openstreetmap.org/index.php/Map_Features there is
a discussion of possible tags for map features. Would it be a good
idea to add defaults for tags where it makes sense. For example
in the absence of the tag "oneway" it would by default be assumed
as no. The possible advantage is less information in the database.
The disadvantage is that clients need to be a little more
intelligent.

There is also no distinction between accepted values and recommanded
values. Again the "oneway" tag is listed as possibly "yes,no,true,false,-1"
Now I assume that yes and true have the same meaning. But as a client
what is the recommanded value to use: yes or true?

2. merging

What are the experiences with multiple people working on the same
area? If I understand correctly other editors try to implement a
sort of live editing.

For the moment I've implemented a smart upload feature that replays
the undo history. The advantage over a true live editing mode is
threefold : much less interaction with the database (if you move
a trackpoint 3 times, only the last position is sent), much smoother
editing and the possibility to continue working in
network-disconnected state.

I feel (but have no idea) that even with true live editing two
people working on the same area at the same time would be painfull
at best. I make
this educated guess from experience with source code distribution
software (cvs, svn, ...).  Has there already been experience with
this kind of scenario?

3. database load

It is wonderfull that the 0.3 API allows one to do a GET map?
on a random area from the client side of things, but strikes me
as a bit odd though. This puts a big burden on the database and
makes that call essentially uncacheable. Has there been any
thought on a call where the earth is divided in fixed boxes
of a a few sq km? As long as nothing changes in that area
the result is perfectly cacheable. This means that looking up
busy (and often also complex) areas would never need to
query the database itself.

If there is some interest I'd be willing to think some more along
those lines and see if it is implementable at all.

cu bart