[Talk-GB] Importing Website Data
tom at compton.nu
Thu Dec 21 09:33:02 UTC 2017
On 21/12/17 09:15, Ilya Zverev wrote:
> Frederik and Tom, please explain what has been wrong with the last import, and why osm_conflate + cf_audit tools used for it (conflation + community validation) still do not attain the required quality for OSM contributions?
I wasn't commenting on any particular import just on the general principles.
I was merely trying to point out that the view that people often have of
"official" data as somehow perfect is often far from the truth. I've
heard enough real world stories of databases inside companies to know
just how far from reality it can be.
People often imagine these things as perfectly curated and fully
normalised and standardised when the reality is often that they're
maintained as an excel spreadsheet by this weeks intern.
> How would you build a process for importing large batches of business chains? Can I improve something in my tools, or should I build something better from scratch?
Well I probably wouldn't because it doesn't especially interest me and I
have no commercial reason for wanting to do so. Plus I know that it's an
extremely hard problem.
It's quite true that an import may well be better than nothing where
things haven't been mapped or aren't being actively maintained, but it's
equally true that an import that includes updating existing objects may
sometimes make things worse, and I don't know how you can tell when you
are making an object worse.
That's what makes it so hard.
That said I will certainly agree that what you're doing is far better
than what many companies trying to get their clients locations into OSM
have done in the past.
Tom Hughes (tom at compton.nu)
More information about the Talk-GB