[OSM-talk] Automated edits code of conduct

tuxayo victor at tuxayo.net
Tue Jul 12 01:03:07 UTC 2016


On 10/07/2016 23:56, Christoph Hormann wrote:
> Remember OSM is largely a do-ocracy - those who put work into developing 
> the rules have a significant influence on their content.  This does not 
> make them illegitimate.  

The questions is how legitimate are they. To know if we can enforce them
strictly.

> Both participating in creating and improving the rules as well as working on the DWG making sure mappers comply with the rules are open to everyone.

Is joining the DWG necessary to enforce these rules/guidelines? If
that's the case then it's to be expected that contributors spotting an
issue don't take the time to join the QA effort and simply call the DWG.
Which in turn is overloaded and can't do it's work with the same level
of quality that the one it's enforcing.

---

On 11/07/2016 00:08, Michael Reichert wrote:
> Both Import Guidelines and Automated Edits Code of Conduct are
> guidelines which will reduce the likelihood that your import/mechanical
> edit gets reverted.

If I understand correctly, this means that they are and should be
enforced very strictly right? Because respecting them only *reduces* the
likelihood of revert.
Then they should be held to high standards for their definition (at
least like tag creation).

And their enforcement should also follow rules. The same way the
enforcement of the tags defined in the wiki (which most automated edits
are) must follow rules.
QA have rules and guidelines and "meta QA" should also.

---

On 11/07/2016 01:08, Frederik Ramm wrote:
[...]
> Because his edits stretched over several days and changesets, and
> because the changeset comments contained no hint at whether or not the
> particular changeset did contain this kind of un-discussed mechanical
> edits, the DWG member executing the revert - that was me - only did a
> cursory inspection and in doing so, reverted a few changesets that were
> *not* mechanical edits.

Was this collateral damage necessary to force the infringing contributor
to respect the AECoC?

Take some of the less blurry(about the automated edit nature) comments
from changesets whose bounding box was not country or worldwide scale:
- fix cycleways
- rivière sous forme de multipolygone
- fusion de chemins
- précision d'une intersection de voies
(some had 3 to 5 occurrences)

They should have been given the benefit of doubt to save time by not
reverting them and avoid highly probable collateral damage.
It the same reason we want automated edits to be regulated, to avoid
collateral damage. Most of the changesets would have been reverted
anyway and message to respect the AECoC would have passed.

If one(DWG member or any other contributor) fixing someone else's
mistake or fault doesn't have the time to be confident to not damage the
database then doubt should win. We are talking about whole changesets,
not a changetset with infringing data mixed mixed with legitimate data.

[...]

> The automated edits code of conduct has been created as a result of DWG
> work, where we often have to deal with the detrimental effects of badly
> planned, badly executed lone-wolf edits.

An executive power making it's own laws does a great damage to the
legitimacy of the said laws.

It needs to undoubtedly reflect the consensus of the community. So non
DWG members could refer to it and enforce it without needing to add
workload to the DWG (as long as account blocking power is not needed).
That would also allow DWG members to intervene with a greater legitimacy
because it would not come from their status.

[...]

> Is it *really* a problem that some rules are not shown to people when
> they sign up? In my opinion, mass edits are an advanced enough topic
> that, if you research it enough, you *will* be pointed to these rules,
> or find them in countless answers on help.openstreetmap.org.

I agree that showing them at sign up wouldn't help. However it's to be
expected that first time mass edits are done without knowing the AECoC
as nothing more than the JOSM search and replace tool is needed. Is not
like importing which require more documentation.

That's why it's important to not blindly enforce (which tend to happen
by lack of time) the AECoC because the final objective to make
contributors willing to do QA do it the right way.

A major (maybe bigger?) issue is DWG's lack of resources. Should the DWG
actually handle all this work? Has it been considered to document the
handling of these particular issues so other contributors could share
the workload?
The reporting of AECoC violations could be done in a dedicated open
mailing list so we could have accountability about how these issues are
handled.
*Any thoughts about this? This is a concrete proposal.*

> I'm all for discussing the rules we have, but I'd like to know what
> exactly the problem is. "There has been no vote on these rules" is not
> the honest reason for this thread

Why? Considering the standards required for tags and automated edits,
not having comparable ones for the content of the AECoC is inconsistent
compared to it's importance.

> and I refuse to be drawn into an insincere, endless procedural
discussion just because someone has an axe
> to grind with DWG.

Why having a past incident with the DWG should devalue someone's
*arguments*? It's not like this topic has been received with an
unanimous denial of the raised issues. So please stick to the topic and
arguments and not on the people. It's already easy enough to become
dispersed due to the broadness of the subject (we already reached
armchair mapping...)

---

On 11/07/2016 01:23, Matthijs Melissen wrote:
> My main issue with the AEcoc is that it is nearly impossible to comply
> with, especially the part that says that community consensus is
> necessary (or rather, "said", because this requirement seems to have
> been silently removed).
>
> Could you point me to a single worldwide mechanical edit that
> satisfies the AEcoc guidelines?

They are broad enough so one opposition would be enough to block and
revert almost any mechanical edit. That's why it's even more problematic
when enforcement is done without the time required to avoid mistakes.

---

On 11/07/2016 01:58, Éric Gillet wrote:
> 2016-07-11 0:08 GMT+02:00 Michael Reichert <nakaner at gmx.net
> <mailto:nakaner at gmx.net>>:
>
>     We don't have a Don't Delete Everything Policy. Nevertheless, we
revert
>     vandalism (if we discover it).
>
>
> But what is vandalism ?

It's decently well defined here
https://wiki.openstreetmap.org/wiki/Vandalism and more importantly there
are response guidelines so any contributor can contribute to handle
these issues and must follow at least some kinds of rules.

---

On 11/07/2016 02:02, Frederik Ramm wrote:
> There might be a potential misunderstanding here; some people seem to
> believe that the policies outlined in the Wiki are some kind of "law"
> and that if you comply with it, you are always "right". (Wikipedia tends
> to run into a "lawyering" problem with this - they have policies, they
> call somebody out for doing something stupid, and the person then says
> "but I have followed all the policies, you cannot do anything, ha ha!".
> This is great fun for those who do stupid things and have a lot of time
> to conduct procedural discussions, and a great nuisance for everybody
> else in the project.)

I see the issue but how can we force the contributors to at least
discuss their automated edits if at least that one rule isn't given a
"law-like" status?

> On 07/11/2016 01:23 AM, Matthijs Melissen wrote:
>> Could you point me to a single worldwide mechanical edit that
>> satisfies the AEcoc guidelines?

> I can't but then we don't track them at DWG - we don't grant
> permissions, we only act when we either hear complaints, or see faulty
> (or otherwise problematic) edits ourselves. Haven't you done something
> about musical instruments once? IIRC there was a bit of an issue with
> you asking for a "vote" on the issue, thereby making it sound as if 51%
> were enough to carry such an edit... but you did run it in the end,
> didn't you?

Can we stay constructive? This example is half an attempt of
discreditation instead an actual argument.

---

On 11/07/2016 10:35, Richard Fairhurst wrote:
> DWG is a committee of the elected OSMF. If you don't like it, you can vote
> for directors on OSMF who share your viewpoint, who can then vote to
> instruct DWG accordingly.
>
> This is called representative democracy.

Is there really any realistic chance of influencing the DWG and the
AECoC that way?

> The alternative is direct
> democracy, where fundamental policies are put to a vote (or "referendum")
> among an ill-informed, over-emotive, easily stirred-up population.
This is a
> really, really bad idea. Trust me on that one. :(

Are the contributors involved in QA (those cleaning the base with
automated edits, those watching the edits to spot errors/vandalism/AECoC
violations and the DWG) as much subject to these issues that voting on
the most important point of the AECoC wouldn't have been possible? We
manage to vote on tags even if take time and the filter if very restrictive.

---

On 11/07/2016 10:36, Christoph Hormann wrote:
> Another thought: maybe it would be helpful to think of the DWG work as
> some kind of fire brigade rather than police.

Due to the conflict prone nature of the rule enforcing work we are
talking about, and the sanction power used in some case. The fire
brigade analogy can't work compared to the police (or justice) one.

> if you find a user doing lots of bogus
> edits with a few correct ones mixed in between you can - no matter if
> you are a normal mapper or DWG - revert those changes in total (after
> trying to talk to the user of course).  There have been in the past a
> few cases where the ratio between factual and bogus edits approaches
> unity and where therefore there has been discussion how to deal with
> that.  But as far as i can see none of the changes mentioned in this
> thread can be considered borderline cases in that regard.

Have you looked at the changesets that are claimed as collateral damage
by Test360?
http://www.openstreetmap.org/changeset/27888534

At least half of them have a comment that would suggest that they are
not mechanical edits and there seem to be all atomic changetsets, not a
mix in one changeset.

---

On 11/07/2016 11:18, Christoph Hormann wrote:
> I'd be somewhat careful here, the OSMF is not really democratic in a
> strict sense (it lacks the typical division into executive and
> legislative bodies for example) and due to the extremely small size of
> the OSMF membership (a few hundred) compared to the size of the active
> OSM community as a whole (25k every month) it is also not really
> representative.
>
> This of course does not mean a direct vote of some sort on things (see
> the tag voting process) would be any more representative.

What in the tag voting process shows that such a process would not be
more representative than the OSMF/DWG one?

---

On 11/07/2016 14:15, Éric Gillet wrote:
> Limiting the automation doesn't necessarily reduce the raw number of
> errors. What it does is that in case of an mapper/software error, the
> error may be applied to less content than a large edit.
> But contributors can put a lot more focus and time in the "automated"
> edit than on each one-by-one manual updates, so I don't think the net
> gain of "automated" edits is negative.

It's not negative only if there are a minimum of rules to avoid lone
wolves doing large scale mistakes. We also want to avoid discouraging QA
work and letting the base rot. (e.g. public transport schema transition)
So it's a cursor that we as a community must set to maximize long term
data quality. That's why the process of setting and enforcing this
cursor is critical and should be more documented and accountable.
(especially for enforcement) Without forgetting the limited resources
available for moderation as in every community.


Cheers,
-- 
tuxayo



More information about the talk mailing list