[Imports] Import Working Group conference call
balrogg at gmail.com
Thu Nov 5 00:06:11 GMT 2009
2009/11/5 Sam Vekemans <acrosscanadatrails at gmail.com>:
> I use a headset & can have video with skpe & txt chat MUCH easier.
> (its also free for everyone :)
Alas forces you to run closed-source software which some people are
not going to do.
> Anyway, i wanted to go over comparing Canada vs. usa import (social impact)
> This is something thats not talked about on email, but VERY LOUD in the subtext.
> This is the direct reason for the approach for the CanVec data im working with.
Here are some notes on the topic and some off-topic notes related to
your other posts about CanVec that went by without much discussion.
One reason the USA import didn't cause all the distruption, and the
main difference between it and the other imports that I notice is that
it was done relatively early in the project's lifetime and finished
timely, so there wasn't much conflicting data. I really wish I could
go back two-three years and do the import we're doing now in Poland,
that would have saved us perhaps months of work. And every subsequent
import is going to need more work. So the quikcer it's done the more
work you save yourself and to any mapper who potentially maps the same
data through surveying or other means, putting their work into it,
while they could be mapping the finer details.
Unfortunately there's also a tradeoff when importing live datasets
(such as CanVec, TIGER and what we're importing here) because the
earlier you do it the more outdated the dataset in OSM database will
be at the moment when actual mappers pop up in the area -- this is
considering we don't have any kind of strategy for updates yet, and
for some datasets there isn't going to be one for technical reasons.
This is one thing I recently noticed -- that like software our data is
bitrotting where not maintained by local mappers. Importing data in
an area you don't visit creates an unmaintained map. While the
original dataset keeps being maintained by its authors (government or
Nevertheless there is also a good reason to wholesale import all the
data available at the given moment if you have enough time to do it,
and it's one thing that I believe hasn't been considered in the CanVec
import strategy yet (the strategy which you mentioned a number of
times where you make the converted data available at a public site and
let local mappers import as they wish). That's how the local import
here was supposed to work too, however each imported piece means a
considerable amount of work (fixing up the tagging & naming,
deduplicating objects already in OSM, and eventually connecting the
two datasets to make one), work that can easily be screwed up if done
in hurry, if done by not-so-skilled users or if done by users who just
ignore the guidelines. I find that if the import process is too easy
because you have prepared nicely sized .osm files that anyone can load
into JOSM and just upload, a lot of users will jump on it even not
having the time to put enough work into it to produce good data, or
even inform others that they've started the import in a given area.
It's much like with the easy access to usenet or some chat networks
(MSN?) where the culture is totally different than say on irc, and
everything points at the ease of access.
That's why I allowed myself to process and import some areas I've
never visited, as much as the time allowed me, and that's also why I
don't supply nicely sized, ready-to-use .osm files now. I just supply
the converter script and the URL to the original repos that you can
take and convert, it's still trivial to do but it's some kind of
technical barrier that hopefully filters out people who have no will
to do the whole task relatively properly -- some of these people are
those who don't care much about the rest of the data in OSM but want
to have a working gps routing in their area without doing the mapping
on their own. The hope here is that the importers will either have
the skills and time to do it properly or write to the local forums
asking for instructions / asking for somebody else to the import, and
likely someone will do it.
More information about the Imports