[OSM-talk] To John Bright (unwayed segments to unclassified ways)

[Peter Miller]

> 
> Message: 3
> Date: Tue, 25 Sep 2007 23:06:14 +0200
> From: Doru-Julian Bugariu <j.bugariu at wad.org>
> Subject: [OSM-talk] To John Bright (unwayed segments to unclassified
> 	ways)
> To: talk at openstreetmap.org
> Message-ID: <46F97846.9050800 at wad.org>
> Content-Type: text/plain; charset="iso-8859-15"
> 
> Hello John,
> hello all.
> 
> It is not helpfull to convert untagged segments to ways of type
> highway=unclassified. Converting segments this way, makes it almost
> impossible to find them after conversion to tag them properly.
> 
> As said before (by me and others) untagged segments are often used to
> mark work in progress. I think, it is better to convert them in a few
> days / weeks, when a consesus about how to tag them (maybe
> conversion=API05 or so) is here, so validator plugin or maplint can
> recognise them properly.
> 
> 
> Greetings,
> Banny

+1 AP105 would be good. (I also really don't want to loose information about
the difference between an incomplete way and unclassified ways). I also want
map lint to pick these incomplete ways out for me.

Thinking about other things that might go wrong, here are a few
questions....

What will happen to ways that are currently reported by map-lint as
'unordered', either because they are linear but made from segments listed in
a random order? or alternatively because they have branches? How will these
be represented as sequences of nodes?

I assume that segment direction will be lost and that the direction of
traffic for one-way streets will defined by the order of nodes in the way
(Segment order seems to have no purpose any more, is that true?)

I would have thought the manual elimination of tags in segments is also a
good task to be doing prior to the translation. I have been doing this in
central London (Hampstead -> Victoria, across to the Isle of Dog and up to
Stratford) but have been waiting patiently for new osmarender rendering of
the area for a couple of days. I have found a number of hidden road names
and other useful tags in the process and have also done a general clean up
removing unconnected nodes, closing ways for buildings, removing a small
about of grafitti etc etc.

For the migration can we do some test conversions of the DB to API05 format
and run them in parallel for a couple of weeks until we have all had a
chance to test the new interface/db format? Possibly we folk a one-off test
version of the DB in API05 format on a test server and try it for about two
weeks while still using API04 for the master live DB during which time we
can check that we are happy with the conversion algorithm and shake out any
problems? When we have confidence we have got it right we will dump the test
api05 DB and convert the master DB to the new format for good. If we don't
do this I think it is very likely we will miss something of importance. I
really think this would be worth the extra work involved to ensure a smooth
transition and avoid the problems we have had in the past. I also don't
think something this simple happens without some significant oversights
however much care is put into the planning in advance.



Regards,



Peter