[Talk-GB] Importing Shell fuel stations

[Mark Goodge]

On 28/12/2017 19:31, Lester Caine wrote:
> Get the return address right ...
> 
> On 28/12/17 16:12, Colin Spiller wrote:
>> I've been adding postcodes in the Bradford BD area using Robert & gregrs
>> useful tools. I've just noticed that the Shell station at the Rooley
>> Lane / Rooley Avenue junction BD5 8JR is now reported as having an
>> incorrect postal unit (the final two letters of the postcode). This
>> postcode appears widely on the internet for this site, but the RM
>> postcode finder thinks it should be Rooley Avenue, BD6 1DA.
> 
> PAF file has ...
> Shell Filling Station
> Rooley Avenue
> BRADFORD
> BD6 1DA
> 
> and BD5 8JR is not listed having been deleted in 2009
> http://checkmypostcode.uk/bd58jr so the real problem is does one leave
> the faulty postcode in place because we can't use the PAF data or do we
> validate postcodes against the codepoint database and remove those that
> are not listed

It's an interesting conundrum, on several levels. We can certainly 
validate against Codepoint Open or the ONSPD, as these are open data. So 
if they say the postcode is impossible (because it's defunct), then we 
can definitely delete it if we want to.

Replacing it with the correct postcode, though, is harder. We'd need a 
source that isn't derived from PAF. But Googling for this particular 
station, all the sources have the old, incorrect postcode - even Google 
itself! (I would expect they're all using the Shell data, of course).

So that leaves us with three options, at least initially:

1. Leave it as is. We know it's wrong, but it's consistent with every 
other source, and it's from the only canonical source.

2. Replace it with the right one. More useful, but potentially risky 
from a licensing perspective.

3. Delete it and leave the entry with no postcode. Probably the best we 
can do as far as accuracy is concerned (in line with the general 
principle that data is better missing than wrong, if it can't be right), 
and avoids any licence conflict. But this is the least useful for users 
of the data (since, in this case, even the wrong postcode will identify 
the location in practice - for obvious reasons, Royal Mail will deliver 
to defunct postcodes long after they have been deleted, and many 
sat-navs will work with defunct postcodes too).

Maybe the best solution is to leave it alone for now, and see if we can 
persuade Shell to fix it. Deleting the postcode risks it being re-added 
by someone else who spots its absence and decides to be helpful, without 
realising that if they use the RM postcode finder to validate it that 
isn't compatible with OSM's licence.

Mark