[Imports] Harmful elements in taginfo tag cleanup process

Ilpo Järvinen ilpo.jarvinen at helsinki.fi
Tue May 5 19:46:34 UTC 2015


On Tue, 5 May 2015, Frederik Ramm wrote:

> On 05/05/2015 03:33 PM, Ilpo Järvinen wrote:
> > Please note that I don't want to discredit all of that process as there's 
> > also many useful edits out of it but unfortunately there are too many bad 
> > apples currently in it too, IMHO.
> > 
> > Can something be done to stop these careless elephants messing the good 
> > work of the mices? 
> 
> "Fixing" stuff is always wrong if it deletes good data.
> 
> Someone who carefully fixes stuff will look at the details and make sure
> they do not delete potentially useful data.
>
> Someone who simply replaces all X tags to Y without looking at the
> individual case is making an automated edit that should be reverted;
> either by the user themselves after you point it out to them, or by a
> concerned members of the community, or by DWG (data at osmfoundation.org)
> if the user is uncooperative.

First of all, I'm more talking about deleting tag X rather than replacing 
X with Y as it seems that around here the tag replacing so far have been 
much better in quality than the deletes (I don't mean the deprecation 
related mass replace changes with this which I see you need to deal often. 
Those are different issue and are IMHO unrelated to the taginfo tagspace 
cleanup efforts). ...Obviously I can only guess how each "fixer" found the 
tags he/she deletes unless that's explicit in the changeset comment but 
it's hard to believe that they would be mostly unrelated to the taginfo 
cleanup process and found by chance alone! The frequency of the delete 
incidents has increased from almost non-existing to around 1 per week 
since the taginfo keyspace cleanup was announced.

We certainly revert those already. However, this was not exactly what
I meant with my question. I'd be more interested to know if there could 
be something that would prevent these in advance. That is, if something 
could be done for the sickness, not just to the symptoms.

What I'm slightly afraid is that we end up missing some of them and don't 
notice them until finding the cause is hard to track down (deleted data 
is quite hard to find after months or so, possible but definately not 
easy unless one can pin-point some object that still exists in DB).

Personally I find these types of deletes pretty similar in process and
their effect to undermine the local community/surveyer effort to that 
Mechanical Edits Policy would seek to protect already. But obviously 
every such fixer I've questioned about this so far have declined that 
their changes fall under MEP as they "review" everything they change. 
However, quite often they still clearly admit on the same time that they 
didn't even understand the meaning of words used in the keys (reasons for 
removal typically are roughly "only few and used locally"). In one case 
one key is already hit twice by a delete. I'm afraid we're doomed to 
revert the same tag deletes once per every month?

The response to comments about these deletes varies from admiting that 
didn't understand it and generally constructive to rather dismissive to 
the local community efforts. Perhaps the taginfo cleanup process would 
need more guidance than there is currently in place as the process itself 
resembles what MEP tries to address but it's borderline whether an 
individual edits being done to non-local context when "reviewed" really 
would fall under MEP or not? On the same time I wouldn't like to see too 
big burden being put on the useful work that is spawning from the very 
same process.


-- 
 i.


ps. I'm starting to sympathize more with DWG effort as I suppose they need 
to deal with this kind of tiring issues almost daily over and over again. 
...So also take this as a thanks to all of you applicable :-).


More information about the Imports mailing list