|
Hi, I work as an OSM-community manager for the Danish Cyclist Federation. We're going all in with OSM for all our map needs and my job is to teach as many of our 17.000 members to edit. We're teaching Potlatch 2, but have problems with hitting the download limit almost instantly during sessions. We're guessing it's because our network uses a proxy with a single outgoing IP for all 10+ computers in use. This is bad news, not least because I have to go on road trips around the country and use unfamiliar network setups. 1) What sort of network setup would be required to avoid hitting the download wall? I'm not much into this, but please feel free to shove tech-stuff at me. I'll pass it on to our resident buffs. 2) I've read earlier posts saying there is no possibility of having an IP-address put on a whitelist. Is this still the case? Regards Andreas |
|
The map data downloads are capped per-IP, so any kind of proxied network (e.g. NAT) is going to have this problem. If you are using a network where the different computers have different public IP addresses, then you will have much less of a issue. There is currently no solution available for per-IP whitelisting, mainly because we don't have the manpower to deal with non-automated solutions like that. There have been discussions about taking into account signed-in users and even users-who-upload-changesets and raising the limits for these people, but again neither of those options are currently active. The best solution is to encourage people to zoom in as far as possible before editing, so that they download the minimum possible data for the area that they are editing. 1
Hi, thanks for the answer. We're always very careful about having people zoom in, but it just isn't enough. Also, we can't change our network setup or bet on finding suitable networks as we go on OSM road trips. I think our best chance is to buy a small heap of USB-modems. They'll hopefully be fast enough and have their own IP's. It's an unfortunate problem, but I understand the reason for the limit. Hope you find a solution for networks, though. Thanks a lot for the great work you are doing:) /A
(03 Jun '11, 12:50)
Hammershoej
1
Thanks for the explanation Andy. This situation impacted a mapping party in Uganda: https://twitter.com/#!/batje/status/76956988948496384 Is there any stats on how often this happens? And any idea on what the manual workload would be? What does it take to add in the tweaks to raise limits?
(04 Jun '11, 16:10)
mikelmaron
Hi Andy. Thanks for the explanation. I was attending the course that Andreas describes. We did not use any kind of proxy server. So in that sense the HTTP requests did not come from the same machine (IP address). We were however sitting behind a NAT, so in some sense we were sharing the same GLOBAL IP address. Can you please elaborate on your comment "If you are using a network where the different computers have different IP addresses, then you will have much less of a issue."? Apparently this was our situation and we still had issues regarding this. Thanks in advance... Jarl
(22 Jun '11, 13:38)
jarl
well, a NAT is a form of proxying in that the NAT machine intercepts the requests, alters the IP addresses and forwards them on. I've updated my answer to specifically state NAT and that it's the number of public IP addresses that is important.
(22 Jun '11, 14:38)
Andy Allan
Would proxying some of your requests via a 3rd party proxy server help in this case? (Or multiple different proxy servers.) I'm not suggesting you try to circumvent the protections on people flooding the API, but it sounds like you simply have multiple people trying to use it normally.
(23 Jun '11, 00:35)
Ebenezer
Hi Andy. Thank you for clearing out the heuristics for determine individual "users", I would like to suggest the heuristics to be improved to distinguish individual users even behind NAT, but I guess that must be somewhere else in a different forum, can you suggest any good starting points for such a discussion? I have plans to arrange some mapping parties but I hesitate as long as this problem exists.
(27 Jun '11, 14:35)
jarl
dev@openstreetmap.org is the best place for development discussions.
(27 Jun '11, 15:40)
Andy Allan
showing 5 of 7
show 2 more comments
|
|
If you were to use JOSM instead you'd be less likely to hit the limit, since it gives more control over what data gets downloaded when. That's probably not a very helpful suggestion because it's a whole new software to learn/teach but perhaps it's an option for you. JOSM is not as difficult for new users as people sometimes suggest. In a group situation it would be easier to get a demo or ask questions on josm but you are asking about P2
(23 Jun '11, 16:30)
andy mackey
|
|
I'm currently working on an improvement to Potlatch 2 that will typically result in it pulling down less data from the server when you pan. (Well, to be strictly accurate, I'm converting some clever code by Matt Amos to ActionScript for Potlatch 2. :) ) It won't be live for a few weeks but it should hopefully make a difference. |
|
One of our members, a keen open-source man, has offered to administer a P2 white-list for situations like mapping parties. He works for Denmark's largest dedicated IT-security company, Dubex ( http://www.dubex.dk/dubex/profile/ + http://www.linkedin.com/company/40332 ) and I'm sure he would be more than qualified to take on the task. Would the P2 team be interested in this? If so I can set up contact. Ok, would the serverguys be interested then? And who are they?
(24 Jun '11, 08:57)
Hammershoej
The server admin guys are listed at http://wiki.openstreetmap.org/wiki/Sysadmin
(24 Jun '11, 19:21)
apmon
|