Too easy to create notes without setting a location

[angoca]

Not sure if the origin is from this page or any fork, but I see many notes in the middle of the Atlantic Ocean for Iranian places.

Some note's id I have closed:

2735935
2723484
3451396
2057596
1762208
1762208
3443951
1840542
3408709

[ArcaneDots]

Thanks for catching the invalid entries, I had a feeling that might be a problem with all of those duplicates.

…

On Sun, Dec 11, 2022, 12:09 AM Andres Gomez Casanova < ***@***.***> wrote: Not sure if the origin is from this page or any fork, but I see many notes in the middle of the Atlantic Ocean for Iranian places. Some note's id I have closed: 2735935 2723484 3451396 2057596 1762208 1762208 3443951 1840542 3408709 — Reply to this email directly, view it on GitHub <#103>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAIGQRFK2ZYZSPV7NAPNLBTWMVOYTANCNFSM6AAAAAAS2XG2DM> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[iandees]

Thanks for pointing these out. I think these are coming from users of onosm.org moving the pins to random spots on the map. It doesn't look like they're geocoding addresses to these places. It might be a spam attempt?

I'm open to suggestions for how to improve this.

[ArcaneDots]

We should be geocoding the new pin location and then verifying that the address is still valid.

[ivanbranco]

Same is happening in Italy at these coordinates: 41.6900000, 12.7100000

[ArcaneDots]

Please check out my progress on PR #116 which is designed to fix this issue. My solution is to limit marker movement in Leaftlet based on the original's search bounding box. Please read the notes and add your comments.

[ArcaneDots]

Just stumbled over the Iranian version of OnOsm.org. Its source code is hosted on OSMIran on Gitlab. It is a copy of the main fork as of early 2022. The biggest difference is that it supports Farsi which is the main language of the region. Even though Right to Left language should be handle correctly by the browser (via HTML), we will need to verify that everything still works as well.

[iandees]

Just stumbled over the Iranian version of OnOsm.org

Good find!

[angoca]

I created an issue in that fork: https://gitlab.com/OSMIran/note-tool/-/issues/3

[ivanbranco]

Doesn't seem to be onosm.osmiran.ir, their notes have the following text:

onosm.osmiran.ir یادداشتی از یک کسب‌وکار ارسال کرد:‬
نام:
تلفن تماس:
وب‌سایت:
شبکه‌های اجتماعی:
ساعات کاری:
دسته‌بندی:
نشانی:

The ones we found have onosm.org, not onosm.osmiran.ir, and all the fields (Website, Phone number ecc.) in english.

Also seems like it's not even used in Iran, of the last 500 notes, 279 were created from onosm.org and only 5 from onosm.osmiran.ir

[ArcaneDots]

I have create PR #116. It is designed to do some basic validation and geo-fencing of the marker. User can do minor adjustments around the initial point. Anything more is based on the bounding box returned by Nominatin during the original search. That is as long the bbox is larger than the default circle. While outside the circle, any coordinates must fall inside the bbox connected to nearest valid point. This seems to prevent the marker from be left in water or other places for that might be odd for a business to be located.

[ivanbranco]

Now also american notes started to pop-up at the same location:

https://www.openstreetmap.org/note/3584982
https://www.openstreetmap.org/note/3584983

Same is happening in Italy at these coordinates: 41.6900000, 12.7100000

[ArcaneDots]

The underling problem is that the website doesn't currently do any type of validation of the marker moves. My current change tries to limit the distance the marker can be adjust to prevent these huge deviations. I'm still not sure why there are so many duplicate notes getting created. For the moment, I need some feedback of my PR #116 which should prevent part of the problem. I would appreciate code reviews from those who are able and user testing otherwise. I am particularly interested in thoughts relating to creating a reasonable bounding box based on the original user search.