[OSM-dev] Minutely Diff Broken UTF-8

[Shaun McDonald]

On 1 Feb 2009, at 06:08, Christopher Schmidt wrote:

> On Sun, Feb 01, 2009 at 04:44:46PM +1100, Brett Henderson wrote:
>> I've manually truncated the broken tag in the relation in the
>> daily/hourly/minute changeset files on the planet server.  If you now
>> have missing data in your database you should be able to reset the
>> timestamp and begin applying changesets including the missing data  
>> again.
>>
>> I've also asked TomH to fix the broken tag in the db but it might  
>> take a
>> while given that it's the weekend.
>
> Shaun mentioned that the reason for this one is that the database  
> allows
> 255 'bytes' -- and in this case, the 255th byte cuts off in the middle
> of a character. This should be fixed in 0.6, which will reject thse
> edits outright.

There are 3 reasons why it will be fixed in 0.6:
1. The database will be reloaded as part of the conversion to get rid  
of any of the UTF-8/Latin-1 connection problem.
2. Rails will have the correct database connection information.
3. Rails will have a validation that will work.

See http://lists.openstreetmap.org/pipermail/dev/2009-February/013811.html 
  for more information.

Shaun