On 07/05/12 10:34, Jonathan Harley wrote:
On 06/05/12 17:22, Andrew M. Bishop wrote:
Andy Street<m...@andystreet.me.uk> writes:
On Fri, 2012-05-04 at 14:32 +0100, Andrew Chadwick wrote:
I'd agree that generic consumers will struggle with highway=path,
designation=* but that is a wider OSM issue and not limited to the
path/footway, etc. debate. Anyone using OSM data should be
pre-processing it to take into account local laws/customs and their
particular use case. For example, you are probably going to come a
cropper if you go around assuming that roads across the globe
without an
explicit maxspeed tag all have the same default value.
As the author of a consumer of OSM data I for one would prefer it if
there was a single set of tags worldwide. In my case the consumer of
the data is Routino a router for OSM data (http://www.routino.org/).
That makes sense - but the question is, should tagging be optimised
for mappers/map editors, or for map consumers, if those things conflict?
My personal opinion is that the biggest risk to OSM's future is if we
don't agree on a subset of tagging rules to be used worldwide. The
idea that there could be a pre-processor to handle local laws and
customs is impractical. There are literally hundreds of regions that
might use their own tagging rules each of which needs to be defined by
a geographical region and list of rules. Each consumer of data then
needs to implement the full set of pre-processor rules.
No; only consumers of data who want worldwide coverage (and who care
about the tags that vary around the world) would have to do that. And
I think that would still be easier than getting mappers worldwide to
conform to a rigid tagging system.
I'm not sure what I think is the biggest risk to OSM's future but I
think attempting to impose an unwieldy system of tags on contributors
is right up there. I think a large part of OSM's success so far is due
to its simplicity and informality.
With a single set of rules a way can be taken from an OSM XML file and
it will be immediately apparent who is permitted to use it. With a
pre-processor it is necessary to take the way from the file, search
through the whole file to find the nodes that are referenced by it,
search through all defined regions to determine which one the nodes
belong to and then apply the selected pre-processor.
One thing that we shouldn't lose sight of is that each item in OSM is
created once and edited a few times by a small number of editors but
used many hundreds of time each day by many dozens of data consumers.
Since the number of times the data is read far exceeds the number of
times the data is written (by orders of magnitude) the complexity
should be in the writing side and not the reading side.
I disagree. Consumers of OSM data should embrace Postel's Law.
Besides, rule-based processing is just CPU cycles. Those are far less
valuable than OSM contributor brain power.
Also, there's no reason data consumers have to use "raw" OSM data.
Someone could post-process OSM to produce dumps that have "normalised"
rights of way information, and publish those files for the benefit of
that subset of consumers who happen to care about rights of way being
consistent around the world. I think that's a much better way to go
than laying down rigid rules for mappers, or running bots that try to
bash OSM into the shape needed by a particular consumer.
+ 1
Mappers are far too precious to lose by making tagging schemes that suit
data consumers and not mappers. OSM has grown partly because free
tagging has allowed the base of tags to grow as people who are
interested in a subject add tags that suit that object. The consensus
over tagging is pretty good, just by good sense and a common purpose.
I am certainly in favour of using tags that everyone agrees with, but
certainly not a restricted list whether that is driven by data
consumers, some committee or wiki editors. Even worse are bots or mass
edits that flatten diversity from the database in the name of
conformity. I view changing someone's carefully chosen tag (not just
typos) to something else as vandalism.
--
Cheers, Chris
user: chillly
_______________________________________________
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb
