Talk:Ordnance Survey Opendata

From OpenStreetMap Wiki
Jump to: navigation, search

Discuss Ordnance Survey Opendata page here:

Ideas (Moved from main page)

These are ideas. Not to be acted upon without consideration. Please see the talk page for further discussion

We don't want a big messy import of TIGER proportions. We could try to use the data as a guide to help improve the map. We'll be working on some tools to help us improve OSM in the very near future and we'd really love your feedback and ideas for what you'd like to see in the tools.

Other suggestions

  • Muki Haklay suggests on talk-gb that we will get most value from this data by extracting just the attributes from it, rather than the geometry. Our geometry is already better, he says.
  • Getting the Postcode data in so that it will appear in Nominatim (or getting Nominatim to use postcode data as secondary source.
  • Using the Land-Form Panorama dataset to increase the contour resolution on OpenCycleMap?
  • It's easter, why not have an OS easter egg hunt!
  • Using the Gazetteer to populate the auto-completion text entry list on Potlatch.
  • Some sort of drag-drop tool to place non-matched Gazetteer entries as points of interest onto a background layer.
  • Import building shapes from Street View or another data set
  • Create a WMS Layer for all of the available data (similar to the Toporama WMS Layer used for the Canada Import process, this will allow for direct copying & ensure the full communities involvement --acrosscanadatrails 01:58, 6 April 2010 (UTC)
  • Feed back all the errors we find, maybe a modified Openstreetbugs installation?


"If we need to import geometries which are not yet in OSM, there's an automatic process described here: it has been used for Brest Metropole mass import"

Removed the above from the page, since imports are another thing entirely and should really wait until much later (preferably never, IMHO). But worth discussion. --Matt 14:27, 1 April 2010 (UTC)

Yep. actually mentioning the source tag is probably a slippery slope. I'm slightly concerned that the buzz is out-of-control and we'll have some crappy import lumped on top of the entire UK data within the next 24 hours! -- Harry Wood 14:39, 1 April 2010 (UTC)


I notice some suggestions on attribution have been added. Probably a bad idea as it may encourage imports or such. Furthermore would it not be better to use source=OS_OpenData rather than source=Ordnance Survey, this is more specific and not ambiguous regarding copyright status. Alex McKee 14:42, 1 April 2010 (UTC)

Heh. Think we made the same comment at the same time. I've tried to de-emphasise that information, or at least emphasise the fact that we need to take it slow and discuss it a little bit -- Harry Wood 15:00, 1 April 2010 (UTC)
Yes, I didn't see your comment above when I loaded the page, just after I had submitted already! Nice work on the page, should be quite reassuring for users dropping in to see the community response and hopefully forestall any reckless imports. Alex McKee 15:14, 1 April 2010 (UTC)

Their licence includes the requirement that users of the data (and any sub-licensed users) add the attribution "Contains Ordnance Survey data (c) Crow copyright and database right 2010." My reading of this is that if we make use of their data, such a statement would have to appear "with" to every published use of the data. It would be relatively easy to tack this on to an attribution page for users of the main OSM website / data. However, surely it also means that anyone publishing a map image (eg on a website or in hard-copy form) whose area includes some OS data would have to add the OS copyright statement under the map (in addition to the OSM attribution). Surely this viral copyright and attribution requirement isn't something we want to have to force users to adhere to. Are users of the OSM data expected to check the source tags in order to find out exactly what attribution they need to add to a particular map image? Is the OS's licence compatible with the proposed ODbL we'll probably be moving to shortly? -- Rjw62 22:21, 5 April 2010 (UTC)

Yeah if your reading of it was correct, it would indeed be unacceptable. But things are rarely that black and white. I'm pretty sure the intention is that people like OSM should feel free to use and re-use. Is there an attribution issue for downstream users? I don't know. It may well be that we can just give credit on our Contributors page, as we have for various government data imports. -- Harry Wood 00:25, 6 April 2010 (UTC)
It seems like quite a plausible reading, and one I suspect we would expect of others if they combined OSM data into other CC-BY-SA datasources too. So it isn't unreasonable, although somewhat impractical. If "banning" attribution to some obscure unlinked to wiki page is legal I don't know, but I think the least we can and (have to) do is make the Contributors page less obscure by adding more prominent links to it. For that purpose, I have recently suggested on Legal Faq discussions to make it a "requirement" to have the words "and contributors" in the attribution statement "(c) OpenStreetMap (and) contributors, CC-BY-SA" link to the contributors wiki page. -- Amm 07:33, 6 April 2010 (UTC)
The final paragraph of the T&C's states "These terms have been aligned to be interoperable with any Creative Commons Attribution 3.0 Licence. This means that you may mix the information with Creative Commons licensed content to create a derivative work that can be distributed under any Creative Commons Attribution 3.0 Licence.". I believe we have other data that we've derived or imported from CC-BY licensed sources. If they don't require a specific attribution (and I'm guessing they don't), the OS data shouldn't.
Imho, the licenses themselves are compatible, but not necessarily the way we and down stream users have interpreted them. It is just that no one has so far sued (or even complained to) OSM about it. So we don't know. There have been quite a few discussions about if we according to CC-BY-SA would have to attribute every single mapper that has contributed (to that part) every time anyone shows OpenStreetMap data. As that is completely impractical, OSM has "decided" to declare it acceptable to attribute "OpenStreetMap and contributors" instead, but it is unclear if that is legal. This is, I believe, one of the reasons why people want a switch away from CC licences to OdBL. So unless we ask OS if it is sufficient to attribute them on a wiki page, I am not sure we can simply assume it is, especially not in its current form without any links to that page. -- Amm 09:26, 6 April 2010 (UTC)

Just how important is it to attribute individual edits of nodes and ways with a source = OS_OpenData_StreetView tag and/or an attribution in the changeset comments? I notice under the Attribution section of this page that it only says that this 'encouraged'. The reason I ask is because I have noticed a lot of mapping being done by one user in Lincolnshire that I am almost certain is being sourced from StreetView tracing but the user is not making any comments on the changesets and as far as I can tell is not adding any source tags to any of the edits either. I just though I would check to see what the situation is before I made any attempt to contact the user. --Dysteleologist 14:10, 14 August 2010 (BST)

License Compatibility with ODbL

My understanding of ODbL is that it allows so-called "produced works" to be released by authors without any licensing restrictions. In particular, this means that a person can release a produced work to the public domain, and others can make use of it without any attribution.

I've been in contact with someone at Ordnance Survey, and their informal view is that this way of side-stepping the attribution requirement is not compatible with the Current OpenData License. So under their current license, we wouldn't be able to make use of any OS OpenData if OSM switched to ODbL. And presumably we'd be unable to carry across any OpenData derived stuff if/when the switch to ODbL occurs.

On the positive side, I got the impression that they're keen to allow use under ODbL, so may be looking in to how they could amend their license (or maybe introduce dual licensing) to accommodate this. Their main concern appeared to be wanting to keep their license simple -- which something compatible with ODbL would have a hard time being.

-- Rjw62 17:29, 29 April 2010 (UTC)

In fact, I think things are slightly worse than this. The new contributor terms mean you have to be able to grant OSMF a specific license to allow them the option of re-license your contributions under an unspecified "free and open" licence at a later data. So under the new terms we won't be able to make use of OS OpenData at all unless Ordnance Survey agrees to grant OSMF that ability. I could see them being persuaded to go for OBdL, but surely their lawyers will balk at the unspecified future license clause. Assuming that as part of the move to ODbL, OSMF wants to have all the data under those terms to allow flexibility in future licensing, I could see us having to remove all OS OpenData tainted information from our database. :-(

-- Rjw62 17:20, 12 May 2010 (UTC)

For the record this seems to have been resolved. See mailing list. --MarkS 21:18, 18 November 2011 (UTC)

Available Data

Added a section describing what's been made available. Kinda for my own sanity, as I read all this back-and-forth in the mailing list but had no idea what any of the names meant! Perhaps add links to the data once the OS site springs back to life Done. --Tiiiim 18:16, 1 April 2010 (UTC)

CodePoint licensing problem

Message from license work group in January 2011 makes it clear that CodePoint "Open" data cannot be put into OpenStreetMap (no importing, no mappers individually adding data from it) There's been no formal update since, though there has been discussion: [1]. It seems there's quite a lot of confusion created by license statements on the Ordance Survey and websites.

User:Chillly was firmly of the view that there is no problem doing what we like with it. He has been working on a display which mashes the data together, and it seems likely that this kind of use is acceptable to all parties. The outcome of the discussion about importing (see below) seems to be that we don't want the data *in* openstreetmap anyway. Mashing it together is the best way for end users to work with CodePoint anyway, so perhaps a direct license incompatibility preventing import is no big deal.

Anyone aware of any updates on this?

-- Harry Wood 12:18, 3 July 2012 (BST)

Should we import Codepoint (UK Postcode centroids)?

This is an open question at the moment and needs more discussions, so please don't import it yet until a rough consensus has been found that is in line with other (worldwide) practices in OSM.

Why we shouldn't import Codepoint

Because it's gazetteer data, not geographic data. Good for going from a postcode to a location, but there's nothing really at that location to survey or verify. Like the GNIS import in the US, it'll just create a bunch of junk points. Use it, sure, but better to collect the real postcodes of real houses on the ground. This loses us nothing, because the Codepoint data is available for any use that the OSM data would be available for. --Matt 01:55, 4 April 2010 (UTC)

If the problem of sporadic building outlines is to be addressed by systematically adding buildings, filling in the missing addresses becomes an good task for new mappers to tackle. It is a lot more difficult to highlight missing addresses when there are imported postcodes scattered across the map. Andrew 12:27, 10 April 2010 (UTC)

Obviously another reason we shouldn't import CodePoint would be the license incompatibility issues state by LWG (see above) -- Harry Wood 12:18, 3 July 2012 (BST)

Why we should import Codepoint

Postcodes are geographic data. They are attributes associated with a geographic position and so fit into the core of what OpenStreetMap data is about. Postcodes are also verifiable on the ground, although partially with difficulty, as e.g. many (commercial) buildings have their postcode in large letters on their door. Other methods of surveying postcodes are from local knowledge or shop receipts and letter heads. So it is no harder to verify than other attributes in the database such as the rest of the addressing data that is in the database. Codepoint is unfortunately only centroids, which isn't ideal, as better would be to have it per building, but it follows the usual way of OpenStreetMap to start with what we currently have and iteratively improve where possible and without it being in the database we can't improve it. So I don't think "a bunch of junk points" is an appropriate description. No more than e.g. all the current 1 million place= nodes, which will hopefully slowly get replaced by admin boundaries. -- Amm 12:35, 5 April 2010 (UTC)

Codepoint is a useful data source for looking up where the centroid is, but (as you well point out) that centroid isn't a geographic entity in its own right and is unverifiable. These centroids are generated from the set of locations sharing that postcode, preserving no information about the shape of that set, so they can't be used for adding postcode data to existing data. As for place nodes, there's the difference between the centre of a place in human terms (usually the high street or a major landmark) and the calculated centroid of the admin boundary, which may be some distance away. A better comparison would be the use of is_in, the arguments against which are well documented elsewhere. --Matt 17:56, 5 April 2010 (UTC)
Hmm, I've been looking at the code-point data for my home town - where I've added most of the address nodes anyway. Most of the time it's *really* easy to work out which particular houses have which post-codes. All of the "centroid" nodes in my home town are on houses alongside ways - even on roads which have bends - they appear not to be centroids in the geometrical sense - more a sort of "median" house in that post-code. I've verified I had the right one on some roads where I know the post-code from someone who lives there. So, I would disagree that they "can't be used for adding postcode data to existing data". Most of the time it's pretty easy. Richard B 12:31, 7 April 2010 (UTC)
The point is that they are not really "attributes associated with a geographic position" at all, they are attributes associated with a set of buildings to which post is delivered. Any geographic location, be it a centroid or a bounding polygon, associated with that is a secondary piece of derived data. TomH 10:39, 6 April 2010 (UTC)

Here's my two pence: It can't hurt. The data is very useful in navigation. For example if you want to go from A to B with either as a postcode, you need this data. If you want to find a node for 'A' which is sensible, choose a nearby road or point which you can get to. If the postcode is too far from any road, then it is unroutable. That's how most mapping systems I've seen do it. And we already have NPE and Free the Postcode data so why not this? (Btw, this is my first comment on the wiki but I am a long time contributor under Tom66 20:46, 8 April 2010 (UTC)

Draft post to talk-gb re StreetView Tiles

The process of creating map tiles from Ordnance Survey StreetView in our required format is now largely complete (zoom 17 tiles not yet available) and can be viewed here:

These tiles can now be used as a backdrop in editors. However before you start editing, please consider the following:

1. Don't assume the OS data is either correct or up to date. Use it as a guide and additional resource for your mapping, not a replacement.

2. Please add source tags to any data you add to OSM from OS StreetView. See the tag suggestions:

3. If you're tempted to map in an area you're not familiar with, contact any local contributors first. Please don't alienate other contributors by treading on their toes. They will know their local area better than anyone else.

4. Look for places that have been mapped using Yahoo! imagery and don't have road names. This is a good place to start.

5. Avoid simply duplicating OS data in blank areas of OSM unless you're familiar with the area and don’t have access to other resources such as a GPS.

6. In May the OS will release a further vector based product which may provide better data for buildings than tracing from StreetView. You might want to hold off tracing buildings until the details are confirmed.

How to make use of these tiles:


Select 'UK: OS StreetView' from the background menu in the options box.


Create a new WMS layer with the following url:

or alternatively use the SlippyMap plugin for which you can find instructions of how to add custom tile sources at with as slippymap.custom_tile_source_1.url


1) In the Tools menu, open the TMS Servers Editor.
2) Add the following:
Name: OS Street View
Server address:
Path: /sv/%1/%2/%3.png
Tile size: 256
Minimum zoom: 6
Maximum zoom: 16
Hit 'apply' and add the new TMS to a map layer.

You can find out more about the OS OpenData products on the OSM wiki here:

blackadder 10:28, 6 April 2010 (UTC)

I use the OSM2GO editor application for Nokia Maemo devices - - would anybody know what settings to put in to that to use as a WMS server ? I tried duplicating both the JSOM & Merkaartor settings and neither of those worked.

star-one 15:52, 22 May 2010 (UTC)

Rights of Way

It seems to me that one vital bit of OS information is the definitive ways for footpaths. Looking in my area, I can see some right of way paths not yet defined, and although I can map the missing bits easily in a few hours, there is an issue of whether it would be better to take an OS source of a definitive map (if such a thing exists electronically) rather than editors expressing their opinion on where they think the definitive way is.

Also, with a time limit of 2020 (if I recall correctly) on getting definitive ways recorded, there is an opportunity in this project to compare the OS source with other ways which have been mapped. IanMSpencer 15:21, 27 April 2010 (UTC)

Rights of way have not been released as part of the OS OpenData package so are not available for us to use. --Richard 16:07, 27 April 2010 (UTC)
OK, thanks. Makes sense, though it is annoying that one very important piece of information has not been made available. I do note that the 1:50000 series maps have walking ways on unlike the StreetView thingies, so it will be interesting to see how good the May release is for checking this information. IanMSpencer 17:22, 28 April 2010 (UTC)
The "May Release" was early, and is now available -- I see it's already made its way onto the mirror ratarsed 12:43, 29 April 2010 (UTC)
The OSM mirror has two corrupt files, as far as I can tell; those for NO & SJ grids (based on downloading in two places, both have CRC issues). Looking at the actual data, RoW are missing, and building outlines wouldn't want to be used en-masse (they're not as detailed as Streetview). Street definition is better than Meridian, but does have gaps in ways where one way passes underneath another - In other words, use Meridian for routing, and VectorMap District for rendering; better still, select the best from both for OSM ;) - ratarsed 18:18, 1 May 2010 (UTC)
I am tracking progress of this here [2], but I have not got anywhere in getting the data released.--TimSC 15:58, 11 August 2010 (BST)
As I understand it, Rights of Way data is not OS data: it is Local Authority owned and maintained under statutory provisions (not a clue which ones). The important thing is that Local Authority Rights of Way data, which is legally definitive, is "public data" according to the official Government definition and there is therefore a presumption that it shall be made freely available (both free of charge and unrestricted). Details are still unclear, but this should be covered by the UK Location initiative, which commits to a geodata portal within 13:12, 22 August 2010 (BST)

Modifying Existing Data

Somebody noticed that their detailed ground surveys were being “corrected” using OS Opendata (in this case Carlton, near Nottingham, and some attributes were removed). Like with all modifications to existing data, can I please stress that people should not just assume their data, including OS Opendata, is more accurate than somebody else’s, especially if that somebody else’s data is from an on‐the‐ground survey?—Sward 12:42, 18 July 2010 (UTC)

Can anybody import boundaries of Scottish councils?

There is NO boundaries of councils at all so I think it's very suitable for automagic import. Pawcio 17:49, 3 October 2010 (BST)

Coastline also could be significantly improved by import

when source is PGS. Pawcio 17:49, 3 October 2010 (BST)

Bing compared to OS VectorMap District

I am an avid user of OSM-Inspector to correct not only my mistakes but of other contributors. I also use Bing imagery for verification purposes. Which of the two Bing or OS VectorMap District are accurate and should have the priority over the other? Check the New Forest using OSM-Inspector for massive OS VectorMap District errors.