Wednesday 18 May 2011

Is this a disused pub, or isn't it?



I regularly ride past this closed pub on my "flat but quiet" 15 mile cycling loop. It strikes me as an attractive building, but a sad sight, and it seems to be on the market for redevelopment. I've not stopped to read the notices, but I assume there is planning permission for a change of use.

The other pub in the same village is now owned by a group of villagers, and seems to be thriving. It's odd how I will pass these things for months before it occurs to me to check whether they are recorded properly on the map. Having checked, I've added the open pub (which was missing), and changed the tagging on this one to show it as disused.

This leads on to the controversy about how best to tag features that are no longer in use. I've decided to mark this one as "amenity=pub, disused=yes", which is one of the common approaches. But that's not the only option, and there are good arguments against the approach I used.

For those who don't follow this stuff, the main problem is that it's as though I'm saying "this is a pub - oh not it's not". Anyone who is very thirsty might stop listening after the first half of the sentence. If I'd said "this WAS a pub" it would be OK. Similarly, somebody who is using the raw data to draw pubs on a map, or provide directions to the nearest pub is normally going to search for things described as a "pub". They will find more than 30,000 in the UK, including this one (unless they listen carefully).

What they probably want to find is pubs that are still in business. So they want to ignore the 70 or so that are already marked as "disused", and a few more that are marked as "closed". They will have to eliminate these explicitly. There are also more than 100 features described as a pub where the name is set to something like "Royal Oak (closed)". There are a few dozen more where there is a note attached (in free text) to the same effect. There are also a variety of less standard ways of indicating the same thing - all against features that are basically marked as being a "pub". The more of these that they handle explicitly, the more accurate their data will be. Any they miss can mislead their users.

The main alternative is to describe these things, not as a "pub", but as a "former pub", "disused pub" or even "dead pub". There are several dozen examples of each of these in the OSM data for the UK. The general approach is fairly common, but the actual values that are used tend to vary quite a bit. This approach has the advantage that none of these will match a simple search just for "pub". So the default behaviour of any software that uses the data is going to be what we would expect most people to intend. On the other hand, if they are mainly interested in pubs that are closed, or all pubs whether they are closed or not, then this data is not going to be a lot of help.

So given the choice, why did I chose to mark this is "amenity=pub, disused=yes", rather than "building=disused_pub" or something similar?

Partly it's because there is a well-established scheme for tagging pubs, and another well-established scheme for tagging things that are disused. Sticking to these keeps the data fairly clean.

I'm also a bit suspicious of advocating ways of tagging that make assumptions about how the data is going to be processed. Who is to say that it is most important to make life easy for people who want to identify active pubs? It's the obvious case, but what about people who are interested in pub history, pub architecture, the number of closed pubs. Or, in checking data quality against some external directory. Or (perhaps more likely) giving directions such as "turn left at the Royal Oak".

I reckon that anyone who seriously wants to extract active pubs from the database is going to find it fairly easy to filter out ones that are disused, as long as the tagging follows some basic principles. And if anyone thinks it is going to be too difficult to ignore features tagged as "disused=yes" then they should expect much bigger problems handling the other variants.

But mainly I've tagged it this way because that's what I see as I am riding past. From a distance I spot a pub, and when I get close I realise that it's disused. Once it has been developed it may look different, but for now, that's what it seems like to me.

As the OSM database gets more rich, and more detailed, and covers a wider variety of objects there are a number of areas where contributors need different forms of tagging to describe subtle differences between similar features. They already have access to a number of different idioms that they can use to express their different perceptions.

Some people have a problem with that. They want to drive out subjectivity by defining explicit data structures in great detail. In some areas this is probably the right approach. Consistency can sometimes be more important than other considerations. But in many areas a more subjective and expressive approach can (and in my view, should) be encouraged.

The arguments for avoiding forms such as "amenity=pub, disused=yes" are understandable, but as a contributor it has the advantage of being easy to understand, and apply in different situations. The form is already widely used. Most importantly it expresses what I see better than the alternatives. Although alternatives are also widely used, they lack the level of consistency that some potential users of the data may need.

This isn't a problem just for pubs, of course. Similar issues arise in OSM with abandoned railway stations, canals, and other amenities.

Nor is it a problem unique to Open Street Map. In the same village there's a house with a painted sign outside that says something like "Church House, Formerly All Saints Church, Now a Private Residence". I wonder why they went to the expense of putting that up?

8 comments:

Paul said...

My thought, for what it's worth, is that your labelling is correct, because it is still a pub, and only stops being a pub when it becomes something else, which it hasn't. A (possibly temporary) change in whether it is being used doesn't change what it is.
But I'm sure someone can think of an example where that logic produces a daft result.
Of course you could always buy it and open it again, that would resolve the issue.

Paul said...

Sorry, another thought. Another advantage of using a separate Disused tag is that you can more easily define for each kind of object whether you want to show disused examples on a map. Disused pubs - probably yes if you are producing a map for navigation, presumably not if you are producing one for planning pub crawls. A&E units probably not, as you don't want someone rushing an emergency to a closed one.

SK53-osm said...

For some time I've been worried that my analysis of pub distribution on OSM was distorted by disused and dead pubs. Your post has prompted me to have a quick check on the figures. In my dataset of 28000+ there were only around 125 either with "(Closed)" in the name, or marked disused=yes.

I remember having a pint in this pub in the late 1990s, and it was more restaurant than pub. I think, like the Stonor Arms, it was a victim of property speculation.

You've also prompted me to check Paley Street and I've added The Bridge House to OSM.

moltonel said...

Strange, to me once a pub is closed it's not a pub anymore, it's just a building. I'm not sure that the history of a POI should be kept anywhere else than in the actual OSM history. So I'd have kept whatever tags still apply (if any), and when the place reopens we can tag it as what it actually is.

In my city center when a place closes down, there's no telling what kind of a place it'll reopen as (or when). I guess some amenity types are more likely to change than others, but where do you draw the line ? What about other tags, like the phone or website (which may still be up to date or not, if it's a restaurant chain for example), when do you decide that the data is oudated ?

Unless you're talking about archaeological or historical sites, I think that "tagging the present" is a cleaner approach.

Harry Wood said...

Comparison of lifecycle concepts

There's several approaches for tagging this data. Dead pubs is of rather minor specialist interest. As far as 99% of people using our map data will be concerned, a pub stops being a pub when it stops doing all the the things which pubs do, such as serving beers. If it's boarded up, it's not (for all intents and purposes) a pub any more.

I feel quite strongly that the disused=yes is a very bad idea. I don't really care what funny little tag the funny little dead pub data collection nutters decide to invent, as long as it is anything but amenity=pub

It's a classic case of having fun mapping versus creating a useful dataset. You've had your fun mapping ...and the dataset just became less useable. Why?

SK53-osm said...

In response to the last two comments: the problem is that amenity=pub is effectively fulfilling two functions (at least in Britain): a place to drink beer, and a significant local navigational landmark.

It is the latter use which causes the problems: removing closed pubs from OSM might satisfy the beer drinkers, but it would reduce it's utility for navigation. I may well give directions to a stranger without any awareness of the (highly changeable) status of a pub.

A corollary of this is that locals often refuse to recognise pubs by their new names (or even realise they have been re-named), a classic example was the Nottingham Knight island which kept this name over 10 or so years when the pub had a different name. In the end the owners gave in and restored the name.

The A40 still has the Target roundabout, even though the pub became a Macdonalds in 1986. I've never been in The Angel, Islington or the World's End in Chelsea, but these are pubs whose names have become localities. There are many more such places throughout the country.

Anyway I've pointed out that this appears to affect less than 0.5% of mapped pubs. I'm sure that there are many more mapped pubs which have closed or disappeared.

Peter Dobratz said...

How about:

name=The Royal Oak
building=yes
(and other addr tags)

Basically, this is the same as an open pub, but without the amenity=pub (and presumably delete the phone number and opening_hours if they were there).

This way, you can still use it as a landmark as it is a building with a name, but people looking for beer will search to the next amenity=pub|restaurant|bar.

gom1 said...

FWIW there are about 52,000 pubs in Great Britain, and about 9,000 have closed in the last ten years. See this. A couple of dozen close every week. So there must be around 20,000 missing from the OSM database (assuming that there aren't too many marked as open that are really closed).