I'd like to respond to Tom's points:
On 26/04/13 10:12, Tom Chance wrote:
I worry a bit about the view put across by Chris Hill, among others,
that we should always assume every tag has been put in place with the
greatest of care and so shouldn't be changed; that little-used tags
are probably valuable, rather than being accidental or unknowing
variations on a more appropriate tag.
My point is not that every tag is carefully chosen, but that some are
and a mess edit will not differentiate . Indeed Jerry started this
thread with the very concern that a tag key being changed reduced the
value of the tagging. A wide-scale or mass edit will blunder through
these with little regard and hence the reason for discussion first.
Contacting editors and discussing mass edits first will throw up these
issues.
Such a position shows little regard for data users, who have to deal
with the resulting mess and inconsistency. As with the wood/forest
example, it means you really can't rely on the difference so you just
have to treat them as interchangeable, all meaning "some trees here".
The building key is treated by most, at the moment, as a catch-all.
There are so many pointless variations on similar meanings that people
treat it as "everything under the building key is basically a
building". Without a lengthy translation matrix drawn up from manually
inspecting TagInfo you can't say much more.
I am sure you, like me, are a data consumer Tom, but I'm surprised at
this attitude. Whenever data is consumed it must be processed and
identified, how else do you tell between a wood and a lake? If you
choose to treat two things as equivalent, such as natural=meadow and
landuse=meadow, the lines of code to use both in the same way are
insignificant compared to the effort to mass edit or manually change
them. Once you have written that few lines of code you have solved the
problem once and for all, no matter how many other occurances get added
to the database. You can deal with this as you load the data, extract
the data, render or otherwise process the data, it doesn't matter, but
as someone who uses OSM data I know this is easy, repeatable, reusable
and provides a permanent solution. Mass editing doesn't; someone may
revert your edit or add a new object with that tag at any time in the
future, making the editing solution fragile to say the least. I don't
have a lengthy translation matrix, I have a few extra lines of code I
can copy to my next project.
I routinely change tagging I find that looks inappropriate in an
effort to make the data more useful. I'm happy to see John Baker
trying to do this. I only wish he had discussed his plans on this list
first, as doing otherwise always raises hackles, and that he took care
to inspect each individual case rather than processing batches of
objects en masse.
You make the data more useful to you, but are you sure you are not
destroying someone else's detail? Would you alter the geometry of an
object just so it looks tidier on your map, even though it does not
reflect its real shape? Why is that different to altering tagging? If
you know the area or survey it and actively decide to add or change tags
to reflect more accurately what is there then that is great, but just
homogenising data is awful and potentially reduces the value of other
contributors' work.
I am fully in favour of editing the data in our database and I actively
encourage it by assisting new mappers as and when I can. I want to see
all kinds of edits of any object type, but those edits should add value
and detail, not reduce it.
--
Cheers, Chris
user: chillly
_______________________________________________
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb