On 7/03/2016 8:25 PM, Colin Smale wrote:
How about limiting the list to product categories, according to an
agreed taxonomy? Particularly categories which help to distinguish one
shop from another. For example supermarkets might be expected to sell
frozen food, but occasionally they might not. So frozen food would be
a candidate for a product category. Some sell electrical goods, but
not all of them. Another candidate for a category I think.
The category level has to be high enough that it will be stable and
not change frequently, but low enough that it has some value in
deciding whether a shop is worth visiting.
Striking this balance is something we are not always good at, as the
discussion easily gets polarised and the outcome (if there is a clear
outcome) is often a design-by-committee Frankenstein's monster of a
tagging scheme... There has been a proposal in the last few days to
redesign our highway tagging. Clearly this is not going to make it
(because we have already a very well-established tagging scheme for
this), but it does illustrate that a somewhat scientific approach to
these tagging taxonomies, including some objective criteria to
classify against, may have great value in the future. So in the
context of shops, there must be an existing taxonomy somewhere that we
could leverage. Possibly within the business itself (how do
Tesco/Walmart classify their shops? What attributes do they store?) or
industry organisations or statistical bodies (anything they provide
statistics on, like "how many shops sell childrens clothes?" is a hint
for a product category).
Well the military have been at it longer than Tesco/Walmart .... the
first level is
https://en.wikipedia.org/wiki/List_of_NATO_Supply_Classification_Groups
Of course they go down to individual items .. even identifying where it
was made, "You can have that screw ... from Denmark or UK, Sir?". But
the first level maybe sufficient? There are a lot of them ... some can
be ignored ... not much call for space vehicles nor guided weapons yet?
I do think there are too many ... and at least at present I would think
most won't get used.
Frozen food does not get mentioned .. if you remove the word 'frozen'
then you are fine... Outback shops in Australia sell bread ... always
frozen as it takes too long to transport it 'fresh'. So food types first
...
The branding thing... If you have a TREK bicycle with a frame failure
then you need a Trek dealer to make the lifetime claim on the damaged
frame. So I do think that has a role to play somewhere, if not under
'sells' then where?
I would suggest that a category-based tagging scheme would be
"OSM-Core" but of course anyone can extend this scheme to include
lower-level data such as individual products as required, while
staying within the core category scheme (i.e. not duplicating it or
competing with it).
//colin
On 2016-03-07 09:45, Warin wrote:
On 7/03/2016 6:56 PM, Frederik Ramm wrote:
Hi,
On 03/07/2016 07:54 AM, Mateusz Konieczny wrote:
Just imagine what would happen if
someone took it upon them to actually list all the products sold at a
supermarket. And update that once a week.
I see no problem with that except that anybody would quickly give up and
overall waste time.
It is possible that some supermarkets would make their list of products
available electronically and thus provide an incentive for a clever
hacker to simply convert that to OSM. Once we say we accept a list of
products for a supermarket, few would make the distinction of "surveyed
list = ok, imported list = not ok"...
Such an automatic import would practically be the only way to keep a
list of products current.
But I see no significant negative effects extending to other people
(shop element would be hard to edit for some time until somebody would
remove outdated data would be the largest problem what is not really
significant).
A typical grocery store would carry about 30.000 different products.
Even assuming it would be possible to encode each in a 15-character
string, the "sells" tag would half a megabyte long - for each store. And
update once a week. If this were done for many stores, the impact on the
size of the planet file, the daily diffs, or even the small download you
make to edit an area, would be noticeable.
That is a good point.
How to do both? Have the detail without too much change/bandwidth?
Guide the use of the 'sells' tag that it should be used for items that are
available and predicted to be available for the next year?
That may reduce frequent updates?
There could be default values that could cover the basics ... like pub sells
beer?
_______________________________________________
Tagging mailing list
Tagging@openstreetmap.org <mailto:Tagging@openstreetmap.org>
https://lists.openstreetmap.org/listinfo/tagging
_______________________________________________
Tagging mailing list
Tagging@openstreetmap.org
https://lists.openstreetmap.org/listinfo/tagging
_______________________________________________
Tagging mailing list
Tagging@openstreetmap.org
https://lists.openstreetmap.org/listinfo/tagging
