Re: [OSM-talk] bot vandalism

2009-10-12 Thread Matt Williams
2009/10/12 Apollinaris Schoell :
> I know it has been discussed before but found a lot of broken data created
> by bot bugbuster and this is important enough to bring it up again.
> There was some lame response but never really an explanation what this bot
> is doing. Some edits are correct others are not.
> There is little information in the wiki that there has been some bug fixes.
> But how can I know if the bot doesn't break my fixed data again.
> Should we allow bots at all? Sure if they are useful and documented. But
> there should be a need to do a quality check and review first before someone
> is allowed to run a bot against the whole planet.
> this bot does so many edits and it's very difficult to even find all the bad
> edits in a flood of edits.
> Is this a candidate for the new block feature to stop the bot until it's
> clearly documented and approved by voting or an other form of agreement?

I guess this is a good topic for
http://lists.openstreetmap.org/pipermail/moderation/. It's worth at
least notifying those on that list of this conver

-- 
Matt Williams
http://milliams.com

___
talk mailing list
talk@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk


Re: [OSM-talk] bot vandalism

2009-10-12 Thread Apollinaris Schoell

On 12 Oct 2009, at 24:05 , Frederik Ramm wrote:

> Apo,
>
>   I don't know about the specific case you are talking about. But on  
> a more general note, I am sure that we will be seeing stricter rules  
> for bots, where instead of *politely asking* for things to be  
> discussed before the bot runs & properly documented, we will in the  
> future *demand* that this is the case and threaten to suspend bot  
> accounts that don't play by the rules, and revert their edits.
>

this is exactly why I raised this again.

> I don't think that there will be an approval process for bots in the  
> near future (simply because of the manpower required for such  
> approval).

we need some process. doesn't have to be formal or same each time.  
Just a minimum > 1 user to make such a risky thing like a bot. If a  
bot user can't prove this was done and documented it will be blocked  
on any user request without waiting for feedback to keep damage under  
control.
I am sure there will be at least 5 volunteers to check the concept and  
code for each bot if it's posted to talk. Still no guarantee but  
fundamental bugs will be fixed before any big damage.
a second step must be a dry run with verification of the changes done.

>
> We might also ask bot users to flag the relevant accounts as "bot  
> accounts", to allow easier filtering on the user side ("I am not  
> interested in changes done by bots").

this should be a requirement. maybe we should even monitor accounts  
for big edits across the planet in short time and block them  
automatically if they are not approved bot accounts.

>
> Bye
> Frederik


___
talk mailing list
talk@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk


Re: [OSM-talk] bot vandalism

2009-10-12 Thread Apollinaris Schoell

On 11 Oct 2009, at 22:37 , Matthias Versen wrote:

> Apollinaris Schoell wrote:
>> I know it has been discussed before but found a lot of broken data
>> created by bot bugbuster and this is important enough to bring it  
>> up again.
>> There was some lame response but never really an explanation what  
>> this
>> bot is doing. Some edits are correct others are not.
>
> Always include examples that are bad (changeset + relation/way/node ).

  wanted to raise the problem of bots in general, don't need help to  
revert or anyone to check the data.

>
>> There is little information in the wiki that there has been some bug
>> fixes. But how can I know if the bot doesn't break my fixed data  
>> again.
>
> What did the owner of the but answered ?
>
>> Should we allow bots at all? Sure if they are useful and  
>> documented. But
>> there should be a need to do a quality check and review first before
>> someone is allowed to run a bot against the whole planet.
>
> Should we allow to register everyone only with a valid email address  
> and
> let him edit the whole planet with JOSM/potlatch/script without  
> quality
> check ?
> This new users are doing the same mistakes I did at the beginning and
> they often break valid data and they sometimes don't respond to emails
> if you have questions unlike the bot owner of that bot.

the difference is no Potlatch or Josm user can edit that fast and  
destructive unless it's some evil vandal. a broken bot can destroy lot  
of data before even anyone detects it. If it's real vandalism we can  
simply revert but a 90% useful bot change is hard to justify for a  
full revert to make sure the 10% are fixed.
>
>> this bot does so many edits and it's very difficult to even find  
>> all the
>> bad edits in a flood of edits.
>
> I have a flood of edits as well :
> http://www.openstreetmap.org/user/Nightdive/edits

a lot but this doesn't qualify for a flood yet. 60k changesets in 8  
months compared to ~1.3m in less than 1.5 months
>
>> Is this a candidate for the new block feature to stop the bot until  
>> it's
>> clearly documented and approved by voting or an other form of  
>> agreement?
>
> Provide examples, try to communicate with the owner of the bot and if
> you fail to find a solution with him and the other people agreeing he
> will be a candidate for a block.

there was some communication but never a detailed explanation and  
documentation. At that time I didn't have an example but John Smith  
had one. The real problem is now to find out what went wrong, why did  
it brake and where do we need to fix it.  But again my main concern is  
that users with best indention but not enough knowledge run bots on  
the whole planet. Every mapper will make mistakes from time to time  
but it's limited. a bot will multiply each mistake. this needs more  
than 2 eyes to review the resiks

>
> Matthias
>
>
> ___
> talk mailing list
> talk@openstreetmap.org
> http://lists.openstreetmap.org/listinfo/talk


___
talk mailing list
talk@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk


Re: [OSM-talk] bot vandalism

2009-10-12 Thread John Smith
2009/10/12 Matthias Versen :
> Should we allow to register everyone only with a valid email address and
> let him edit the whole planet with JOSM/potlatch/script without quality
> check ?
> This new users are doing the same mistakes I did at the beginning and
> they often break valid data and they sometimes don't respond to emails
> if you have questions unlike the bot owner of that bot.



> I have a flood of edits as well :
> http://www.openstreetmap.org/user/Nightdive/edits

The difference between a prolific user and a bot is depending on how
the bot is coded it has the potential to do a lot more damage if left
unchecked, in the case of a new user they are more likely to check the
results (rendering) of their edits and adjust or correct their work
before making further mistakes, bots don't have this feed back loop
once they're let loose and the results could be very devastating.

___
talk mailing list
talk@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk


Re: [OSM-talk] bot vandalism

2009-10-12 Thread Frederik Ramm
Apo,

I don't know about the specific case you are talking about. But on a 
more general note, I am sure that we will be seeing stricter rules for 
bots, where instead of *politely asking* for things to be discussed 
before the bot runs & properly documented, we will in the future 
*demand* that this is the case and threaten to suspend bot accounts that 
don't play by the rules, and revert their edits.

I don't think that there will be an approval process for bots in the 
near future (simply because of the manpower required for such approval).

We might also ask bot users to flag the relevant accounts as "bot 
accounts", to allow easier filtering on the user side ("I am not 
interested in changes done by bots").

Bye
Frederik

___
talk mailing list
talk@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk


Re: [OSM-talk] bot vandalism

2009-10-11 Thread Matthias Versen
Apollinaris Schoell wrote:
> I know it has been discussed before but found a lot of broken data
> created by bot bugbuster and this is important enough to bring it up again.
> There was some lame response but never really an explanation what this
> bot is doing. Some edits are correct others are not.

Always include examples that are bad (changeset + relation/way/node ).

> There is little information in the wiki that there has been some bug
> fixes. But how can I know if the bot doesn't break my fixed data again.

What did the owner of the but answered ?

> Should we allow bots at all? Sure if they are useful and documented. But
> there should be a need to do a quality check and review first before
> someone is allowed to run a bot against the whole planet.

Should we allow to register everyone only with a valid email address and 
let him edit the whole planet with JOSM/potlatch/script without quality 
check ?
This new users are doing the same mistakes I did at the beginning and 
they often break valid data and they sometimes don't respond to emails 
if you have questions unlike the bot owner of that bot.

> this bot does so many edits and it's very difficult to even find all the
> bad edits in a flood of edits.

I have a flood of edits as well :
http://www.openstreetmap.org/user/Nightdive/edits

> Is this a candidate for the new block feature to stop the bot until it's
> clearly documented and approved by voting or an other form of agreement?

Provide examples, try to communicate with the owner of the bot and if 
you fail to find a solution with him and the other people agreeing he 
will be a candidate for a block.

Matthias


___
talk mailing list
talk@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk


[OSM-talk] bot vandalism

2009-10-11 Thread Apollinaris Schoell
I know it has been discussed before but found a lot of broken data created
by bot bugbuster and this is important enough to bring it up again.
There was some lame response but never really an explanation what this bot
is doing. Some edits are correct others are not.
There is little information in the wiki that there has been some bug fixes.
But how can I know if the bot doesn't break my fixed data again.
Should we allow bots at all? Sure if they are useful and documented. But
there should be a need to do a quality check and review first before someone
is allowed to run a bot against the whole planet.
this bot does so many edits and it's very difficult to even find all the bad
edits in a flood of edits.
Is this a candidate for the new block feature to stop the bot until it's
clearly documented and approved by voting or an other form of agreement?


--
apo
___
talk mailing list
talk@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk