[HOT] Buildings and residential areas: Buldings as nodes (Q1)

Jan Martinec jan at martinec.name
Mon May 22 20:37:18 UTC 2017


Of course this is not the place for deciding for OSM at large. Note that
HOTOSM's mapping is well within the general OSM conventions - surely it is
decidable "it seems that both is possible, so we could choose either for
our purposes." As for differences between TI and wiki - yes, some
conventions do emerge spontaneously and are codified ex post. Although this
makes the data harder to parse, it lends OSM a certain flexibility.

As for scale, the entire OSM planet file with complete history takes ~60
GB: one flash disk worth of data. It is still a considerable amount of
data, and in the past may have been the largest public database in pgSQL,
but it's pretty far from bleeding edge these days: anything resembling Big
Data is many orders of magnitude bigger today (petabytes and up - literally
a million times larger). For a current extract of one country, OSM tops out
at low hundreds of MBs, for less mapped countries it's rather tens of MBs.
(Of course, if you are using .osm files as baseline, you'll get far larger
numbers - because that's the most verbose format available for representing
the data: a dataset of 33 MB in .pbf has 1.8 GB in .osm format).

I do agree with you with regard to bandwidth; that could be an actual
bottleneck.

Cheers,
Jan

Dne 22. 5. 2017 21:44 napsal uživatel "john whelan" <jwhelan0112 at gmail.com>:

But you can't decide something about OSM on the HOT mailing list.  You can
confer and edit the wiki but that may or may not be followed.  There are a
number of instances where what is suggested in the wiki and taginfo are
quite different.

Data compression was mentioned purely because of the writer showing up as a
computer science / engineer person.  Having worked in computers including
setting standards for some years cost is something they don't always
consider and it is a major part of engineering.  No matter what compression
system is used four nodes will always take up four times the space as one
node.  Maybe not with .7z compression looking for strings in the long lat
but its a good rule of thumb.  Again OSM is now running the largest
database known in whatever it is running in, I forget the name.  It's
really big which means bleeding edge as we used to call it for backups, for
data retrieval etc..  Not quite where you want to be for reliability.

I've seen some areas / projects where building outlines are mapped
accurately.  Ottawa, Canada is one, the data was imported from local
government sources.  Lusaka, Zambia, they used motivated experienced GIS
people but having said I've done a lot of validation and accurately mapped
buildings are not infrequent on a HOT project.

The biggest users of nodes for buildings on HOT projects are experienced
OSM mappers.  They will switch to ways when you give them feedback that
that is what HOT prefers but I think they are the largest source.  The HOT
training group has done a very good job on training HOT mappers to draw
ways.

Cheerio John



On 22 May 2017 at 14:16, Jan Martinec <jan at martinec.name> wrote:

> Hello all,
>
> There's really no "their" where OSM is concerned - the database is made by
> individual contributors, without centralized oversight, and much of the
> mapping is by convention. But looking at the OSM wiki gives some fairly
> strong recommendations: http://wiki.openstreetmap.org/wiki/Key:building -
> to paraphrase, "in absence of better data, building-as-a-node is better
> than nothing at all, until it can be mapped as an area from a better
> source."
>
> As far as "accurate mapping" - that's a problem and a driving force for
> the whole of OSM: have you seen the map in 2010? There was a very similar
> situation worldwide, with most buildings existing as rough outlines or not
> at all; requiring "perfect or nothing" would have resulted in no OSM,
> period.
>
> That said, the quality of mapping does depend on feedback, especially with
> new mappers: is the project with badly mapped buildings representative of
> HOT projects? The ones I've seen (and edited) seem to match building shapes
> and sizes well - e.g. http://tasks.hotosm.org/project/2520 - but that's
> also an issue of task validation, in addition to mapper training; I do
> agree that takes some additional human power in return for much better map
> result.
>
> So, I guess that where time is critical (or space, even though XML
> compresses well and data for mobile apps is compressed even more
> efficiently), buildings-as-nodes are an acceptable interim solution, but
> the ideal case is buildings as well-outlined ways (which is what
> "experienced OSM mappers" do prefer, btw; no retraining needed).
>
> Cheers,
> Jan "Piskvor" Martinec,
> OSM and HOTOSM mapper
>
> Dne 22. 5. 2017 17:54 napsal uživatel "john whelan" <jwhelan0112 at gmail.com
> >:
>
>> But from your computer science background you should realise there are
>> costs involved.  To mark a building as a node is one line in the database.
>> As a way well there are four nodes for a start each with its lat and long,
>> then you have the connecting way.  Have you saved a bit of .OSM and opened
>> it in Notepad++?  Try it sometime.  Open JOSM download a tiny area ie a
>> building and take a look.  Download a node and take another look.
>>
>> So we have time costs in mapping, plus internet costs in uploading the
>> additional information, storage in the main OpenStreetMap database,
>> additional costs in downloading, more storage required on smartphones in
>> the field more processing required at all stages. The time costs in mapping
>> mean given the number of mappers we have few projects will get mapped. We
>> also have experienced OpenStreetMap mappers mapping who may not follow HOT
>> guidelines.  These will need retraining and how will you reach them?
>>
>> In Africa internet transmission costs are much higher than locally in
>> North America so ideally we want to minimise these.
>>
>> Then you get to the added value.
>>
>> Take a look at http://tasks.hotosm.org/project/2656#task/102 when I
>> validated it recently for a highways project it seemed to me that most
>> buildings=yes were twice or three times the size of the building sometimes
>> covering more than one building, at least they were mostly square.  For HOT
>> projects this is not untypical but strange shapes tagged building=yes
>> abound. I've seen tiles when only half the buildings have been mapped but
>> the tile marked done.
>>
>> The true added value is being able to estimate population. How many
>> people are there that need to be vaccinated.  If the buildings are mapped
>> accurately then you stand a chance.  You may have seen some references to
>> the JOSM building_tool plugin, try it if you haven't.  You take the number
>> of buildings and their combined area and you can make some reasonable
>> guesses. The area data is so unreliable you might as well have asked for
>> nodes and to be honest you stand a better chance of them all being mapped.
>>
>> I take it you took all these points into consideration before saying that
>> building=yes nodes should not be used?
>>
>> The place to raise the issue is with OpenStreetMap, its their map and
>> there will be different points of view and it might be worth checking how
>> many there are in the map already.
>>
>> Cheerio John
>>
>> On 22 May 2017 at 11:18, Enock Seth Nyamador <kwadzo459 at gmail.com> wrote:
>>
>>> On personal level I feel mapping buildings as nodes is very wrong so I
>>> avoid it.
>>>
>>> But I think tagging nodes as building should be looked at very well. I
>>> will recommend it is deprecated.
>>>
>>> Best,
>>>
>>> - Enock
>>>
>>> 2017-05-20 11:38 GMT+00:00 Vao Matua <vaomatua at gmail.com>:
>>>
>>>> My opinion is that buildings should be mapped as areas.
>>>> In un-mapped areas it would be best to create landuse=residential areas
>>>> first rather than quickly tagging buildings with single nodes. When it
>>>> comes time to trace buildings it is troublesome to convert single nodes to
>>>> polygons.
>>>> For existing single node areas they could be cleaned up on an as-needed
>>>> basis.
>>>>
>>>> Emmor
>>>> (Palolo)
>>>>
>>>> On Sat, May 20, 2017 at 2:10 PM, Bjoern Hassler <bjohas+mw at gmail.com>
>>>> wrote:
>>>>
>>>>> Dear both, dear all,
>>>>>
>>>>> If we're agreed that it's better to map buildings as areas - should we
>>>>> try to re-map node-buildings from previous campaigns as areas? Or shall we
>>>>> just leave it?
>>>>>
>>>>> Many thanks,
>>>>> Bjoern
>>>>>
>>>>> On 19 May 2017 at 17:54, Cascafico Giovanni <cascafico at gmail.com>
>>>>> wrote:
>>>>>
>>>>>> AFAIK Osmand renders addr: housenumber also for polygons [1].
>>>>>>
>>>>>> Anyway I agree to the general rule that rendering has to adhere to
>>>>>> mapping (and not VS).
>>>>>>
>>>>>> [1] cascafico.altervista.org/public/Screenshot_2017-05-19-18-40-
>>>>>> 41.png
>>>>>>
>>>>>> _______________________________________________
>>>>>> HOT mailing list
>>>>>> HOT at openstreetmap.org
>>>>>> https://lists.openstreetmap.org/listinfo/hot
>>>>>>
>>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> HOT mailing list
>>>>> HOT at openstreetmap.org
>>>>> https://lists.openstreetmap.org/listinfo/hot
>>>>>
>>>>>
>>>>
>>>> _______________________________________________
>>>> HOT mailing list
>>>> HOT at openstreetmap.org
>>>> https://lists.openstreetmap.org/listinfo/hot
>>>>
>>>>
>>>
>>> _______________________________________________
>>> HOT mailing list
>>> HOT at openstreetmap.org
>>> https://lists.openstreetmap.org/listinfo/hot
>>>
>>>
>>
>> _______________________________________________
>> HOT mailing list
>> HOT at openstreetmap.org
>> https://lists.openstreetmap.org/listinfo/hot
>>
>>
> _______________________________________________
> HOT mailing list
> HOT at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/hot
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/hot/attachments/20170522/cf784def/attachment.html>


More information about the HOT mailing list