BTW, if tile server access would be blocked to OSM community and everyone else, except osm.org…
Then direct costs of keeping it running for osm.org alone would likely be greater than current costs, as we would likely lose various donated resources.
(this is not even counting reputational costs, lost marketing, imposing costs on community use, say JOSM download menu would need new sponsor)
imho:
OSM tile servers - motivating more people to join and help. They act as the primary marketing tool. → and more contributors lead to more accurate and up-to-date maps, enhancing data quality for everyone.
Easy-to-use map visuals inspire people to share their geographic knowledge, adding more open data.
( imho: we need more thematic layers in the osm.org )
OSM tile servers offer free access to map data, enabling the development of geospatial tools and strengthening open-source projects like QGIS.
We went over the numbers last time Steve viewed this as an issue. The 2024 budget cost for the standard tile layer was €10,288.59. The 2025 cost is likely to be about €5,000. Restricting tile access to osm.org would likely result in no cost savings, and potentially an increase. Currently the credits for one of the sponsored tile servers also cover other tile layer activities like logging, and if we restricted tile access to osm.org we would be less likely to get sponsorships in the future.
We have upcoming sponsorships that may reduce the 2025 cost to about €2000.
On the time to manage the service, on the ops side the largest cost has been fixing bugs that the vandalism earlier this year exposed. We would have had to do this if tiles were from osm.org only. We have some ongoing work to move the Fastly config to OpenTofu as part of IaC work.
I looked at our minutes from this year. The work related to the standard tile layer service have been
Outside of ops, the biggest use of volunteer time I know of has been the DWG responding to questions about vandalism. This would likely reduce if the standard tile layer were restricted to osm.org only.
The biggest time cost has been repeatedly explaining to the same user that their data on the cost of the standard tile layer is incorrect.
Just for the record: I didn’t say a single word about being free or the largest tile server or 2nd largest or anything like that.
The only way it doesn’t cost anything is because you’re abusing donors.
Someone, somewhere, is paying for this.
Who is paying for the unmetered bandwidth? The sponsored servers? Why are you so reckless with their donations?
If you’re happy to abuse donors to run “free tiles for everyone”, what else are you happy to do?
That’s the deeper point here, the complete lack of discipline or responsibility.
This is really helpful detail - thank you!
I should add that I think Grant is amazing, this isn’t personal. But this continuing idea that “we get it for free so it doesn’t matter” and the apparent lack of oversight is not ok.
Why you are assuming that they are unaware or unhappy how these resources are used? Tile server CDN was donated to be tile server CDN, it would be very surprising if they are unaware that their CDN is in fact used as CDN.
I assure you that noone was mislead, forced or defrauded into sponsoring tile servers. It especially applies to major sponsors of bandwidth, servers and AWS credits.
I appreciate this - as I do all of your writing - you make things really clear.
My own interest here is fairly narrow. I tend to object when folks (not you) minimize the costs, because it can sometimes minimize the “value.” The value of the tile service is high, and I would love for us to find a way to fundraise off of it. If we could put an effective “call to give” or collect emails to which we could send an occasional appeal, or something similar, I suspect it could be significant. The audience is so large, it might be enough that we wouldn’t have to do any other fundraising.
@Mateusz_Konieczny I would love to see a letter from the donors signing off that 80% of their traffic they donate to OSM actually goes to large corporations, porn sites, governments and so on. That would be fantastic. I would frame it on my wall as a reminder of what a wonderful world I live in. 
They are also sponsoring for example Tor, so they are clearly aware that sometimes you also help also not only primary intended beneficiaries.
(Tor in particular has also far more odious beneficiaries in addition to censorship-evading in countries with evil regimes and other nice uses)
Feel free to consider it as complete lack of whatever, but they are not donating your resources but own.
(I just hope you will not try to sabotage this donations, as mentioned OSMF does not need more tricky challenges, we have enough problems to handle with current resources and we do not need more tricky to solve problems as an interesting challenge)
And I think that it this thread demonstrated that in AD 2024 budget year primary cost for OSMF is discussing whether OSMF should use it’s resources on tile servers and explaining than fungible OSMF costs spend on tile servers are minimal? Which was initial claimed concern? This goalpost moving is starting to be confusing.
Note that resources spend on sorting tile users into icky, nice, well liked, too profitable and U-shaped and into whatever else categories would cost more than current general tile availability.
BTW, if these resources would not be donated then it would go to large corporation anyway.
If we limited OSMF tiles to osm.org, we’d be creating a massive disadvantage to mappers due to losing access to those tiles in editors outside of the OSM hosted version of iD, as well as QA tools. Both of those highly benefit from incredibly up to date tiles, and I suspect any free tile service (I’ve only ever seen one) would likely not be re-rendering their tiles on the frequency that our tile servers do. We’d also be impacting those who use other tools like QGIS for OSM work.
If you’re in favour of restricting access to OSMF tiles, how would you handle that? And, how would you assist in implementing that, considering that there’s features that would benefit mappers that have yet to be implemented (see the recent discussions on changeset comments), and from what can be seen, we aren’t currently having any issues due to access to those tiles being available to all (provided attribution is given and fair use is followed).
I’d like to see that too, but I’m not sure if there’s any great ways aside from mentioning it in existing calls. Some of the larger external users like QGIS already mention OSM, but the long tail of users is very long.
Thanks for doing the math.
Certainly these numbers are trivial for an organization the size of OSMF. It would be perhaps a useful exercise to explore what tile serving would cost in the absence of donated services, and whether we think those in-kind donations are replacing monetary ones. But, just because a company donated $X in bandwidth, doesn’t mean they would have otherwise donated $X in cash.
The one thing that has always bothered me about our tile serving is that we don’t set a specific number as the upper bound. I’m not sure how often we cut someone off for legitimate usage in a popular site, but surely we can find some number such that any user that exceeds it is surely in a position to pay for their own rendering. But, if we’re only spending a few thousand euro, it seems like this is a moot point.
Yeah, I don’t see Fastly giving us the equivalent amount in cash, and Amazon might give us other services but even then, I suspect it would be less value to us.
Not really. Note that costs of in-kind donations of services etc are in general lower to donor than any of
For example we may get 1000 dollars in versus for cloud thingy that would cost 900 dollars as a commercial service… While may cost 300 dollars to provide for donor.
At the same time donor may have some benefits for them, that cost us nothing (they can truthfully consider themselves as supporting osm project, make employees happier, demonstrate value of their core service etc). Let’s say that in this toy example it is valued at 350 dollars.
So they would at most donate much less - in this toy example it would be something better 300 and 0. And even in this theoretical toy example it is hard to guess this, as it would depend on their motivations.
But this donations are often at least in large part if not completely motivated by desire to help specifically tile rendering and delivery so it is even less fungible.
(Note: yes, this is educated guessing but very likely directionally accurate, getting accurate values would require espionage that would be costly, unethical, risky, stupid and maybe also illegal)
We have had someone explicitly say they won’t donate any cash but will donate about 40k of services. That wasn’t for the tile layer, but the same will apply to the donors there.
We used to have vague language about heavy usage but we removed it as we’ve never had a problem when users were otherwise following the terms. If some site does come along that causes a problem we can contact them and let them know their usage needs to change. In practice, any site that is big enough to cause load problems would want services we don’t provide, like having a contract with SLAs.
I don’t think the economics work out without a major investment of effort, and even this it’s highly risky. If you start charging, you need infrastructure to manage subscriptions – payments, authorization tokens. You need terms of service, SLAs. You need to respond to customer inquiries. All of this requiring staffing up, so you need to manage that HR. It’s basically starting a small business. Meanwhile, OSMF attention is focused on this rather than on all the other things we need.
This.
The osm.org tiles are old technology (raster tiles) at low quality (1x resolution) to a cartographic design that, shall we say, is… not entirely aligned with current market expectations?
There are already plenty of map tile providers with low pricing. There’s not a lot of room in the market to undercut them and still make a worthwhile return.
So basically all an osm.org commercial tile service would bring to the market is brand recognition. That’s not nothing but I struggle to see it as a particularly compelling offering.
Instead, OSMF can provide tiles for mapper feedback, small projects and prototyping, entirely in line with OSM’s original vision of encouraging map use in “creative, productive, or unexpected ways”, and discourage heavy use by rate-limiting. Which is exactly what happens. There are always opportunities to crank the rate-limiting up and I would be surprised if the sysadmins didn’t keep this under review.
The answers here are pretty definitive, and that’s great, it sounds like you like the way things work.
But not for nothing, the assumptions and statements being made in this thread about branding and fundraising are about as accurate and on point as if I explained to you all how raster files work. I am a current head of brand, and I was on a team that raised a literal billion dollars for a university, and none of you, quite frankly, know what you are talking about.
I say that with humor and also with a shake of the head. At least I frame my inputs and ideas as questions, and I do this sincerely. You all just seem to know everything about everything. So good. It’s been made clear there’s nothing left for anyone here to learn.