[CITE-Forum] reopnened: BBOX with zero extent

Sebastian Goerke goerke at lat-lon.de
Wed Apr 10 03:21:23 EDT 2013

Hash: SHA1

Hi Arnulf,

regarding a single point, you are right. But as said before, within
WMS spec this does not matter if the spec is only regarding layers
here, without considering their content's geometries.

Regarding compliance, it makes sense to have tests like the one we are
discussing, which are looking for the bbox generation of an
implementation. And if it is the case, that there is a zero bounding
box then, the implementation is not compliant to the specs
requirements, no matter if the layers content is a single point or
anything else.
To make it a bit more clearer. You can define a layer with no data but
a bounding box of your choice. This is compliant to the spec and there
could be a use case like having a transactional process behind that to
fill the datasource of the layer for example by users. Nevertheless
the layer is initiated with no data but a valid bounding box.

Best Regards


- -- 
l a t / l o n  GmbH
Aennchenstrasse 19               53177 Bonn, Germany
phone ++49 +228 18496-0          fax ++49 +228 18496-29
http://www.lat-lon.de            http://www.deegree.org

Am 10.04.2013 00:10, schrieb Arnulf Christl:
> Sebastian, right on, I completely agree. But this still does not
> really help to pragmatically solve the issue when this thing is
> going to happen. And if we can think it up it will happen, so why
> not address it right away?
> A point is a point and that means it will have a zero sized BBOX.
> It is no use to forbid this because there will be points and it is
> their perfectly natural right to have a zero sized BBOX. :-)
> My point is that we should simply add a note letting people know
> that this thing can happen (and therefore will happen, as I have
> just recently proven). We are even inviting this to happen with a
> layer that has just one single point in the CITE dataset. There is
> simply no "right" or "correct" way to deal with it given the
> current standard and test process.
> Plus - in my personal opinion - it does make sense to disallow
> things that are perfectly natural, like a zero sized BBOX for a
> point. But I don't really care enough about it to carry it through
> as a CR for the WMS spec, I guess there are more important things
> to worry about.
> Cheers, Arnulf
> On 04/09/2013 10:20 PM, goerke at lat-lon.de wrote:
>> Hi Arnulf,
>> I disagree. I think, WMS SWG was aware of the problem of layers 
>> consisting of only 1 point while creating the standard and 
>> therefore specified that bounding boxes must have an extent >0.
>> My second point is, that following your suggestion makes it
>> impossible to query the bounding box of the defined layer. But
>> the strongest argument is, that in the context of WMS services
>> you never speak about bounding boxes of geometries. It is all
>> about layers. There is no definition, that the bbox of a layer
>> containing only a single point has to be a zero area. So the
>> bounding box of a layer is not restricted to the bounding box of
>> its contents. This makes sense as a service provider could
>> restrict the bounding box of a layer to only a smaller extent of
>> its datasource behind.
>> Following this, the WMS spec is well defined regarding bounding 
>> boxes for layers which contain only a single point.
>> Best Regards
>> Sebastian
>> Hi, instead I would suggest to modify the standard and add an 
>> informative section:
>> 1. "A Bounding Box should not have zero area." Reasoning: A
>> point geometry has by definition a zero extent (area), anything
>> else would be an arbitrary.
>> Then add something like this:
>> Informative note: Point geometries have by definition a zero
>> extent and software has to make sure this does not lead to
>> division by zero (or whatever else tech note might help to
>> clarify the point).
>> Reasoning: This will make sure that implementers look into the 
>> problem and make sure their software handles the problem
>> correctly by at least producing a sensible error message.
>> Cheers, Arnulf
>> On 04/09/2013 07:05 PM, Richard Martell wrote:
>>>>> Hi all,
>>>>> Clause 6.7.4 in ISO 19128 (WMS 1.3) states that:
>>>>> "A Bounding Box shall not have zero area."
>>>>> This constraints applies to both service metadata and 
>>>>> requests. If the current test suite does not verify this
>>>>> for the capabilities document (getcapabilities.xml doesn't
>>>>> appear to), then a new test is warranted.
>>>>> -- Richard
>>>>>> -----Original Message----- From: 
>>>>>> cite-forum-bounces+rmartell=galdosinc.com at lists.opengeospatial
>>>>>> [mailto:cite-forum-bounces+rmartell=galdosinc.com at lists.openge
ospatial.org] On Behalf Of Sebastian Goerke Sent: Tuesday, 02
>>>>>> April, 2013 01:02 To: cite-forum at lists.opengeospatial.org
>>>>>>  Subject: Re: [CITE-Forum] Solved
>>>>> Hi Arnulf,
>>>>> the WMS spec says in BBOX:
>>>>> If a request contains an invalid BBOX (e.g. one whose
>>>>> minimum X is greater than or equal to the maximum X, or
>>>>> whose minimum Y is greater than or equal to the maximum Y)
>>>>> the server shall throw a service exception.
>>>>> This implicates, that requesting a bbox with equal ll and
>>>>> ur is impossible for compliant WMS implementations. As a 
>>>>> consequence, there is somehow an implicit requirement for
>>>>> WMS servers, not to advertise such bounding boxes as they
>>>>> are not useable. Maybe this is something for a change
>>>>> request.
>>>>> Regards
>>>>> Sebastian
>>>>>> _______________________________________________
>>>>>> CITE-Forum mailing list
>>>>>> CITE-Forum at lists.opengeospatial.org 
>>>>>> https://lists.opengeospatial.org/mailman/listinfo/cite-forum
_______________________________________________ CITE-Forum mailing
>>>>> list CITE-Forum at lists.opengeospatial.org 
>>>>> https://lists.opengeospatial.org/mailman/listinfo/cite-forum
>>> _______________________________________________ CITE-Forum 
>>> mailing list CITE-Forum at lists.opengeospatial.org 
>>> https://lists.opengeospatial.org/mailman/listinfo/cite-forum
> _______________________________________________ CITE-Forum mailing
> list CITE-Forum at lists.opengeospatial.org 
> https://lists.opengeospatial.org/mailman/listinfo/cite-forum
Version: GnuPG v1.4.11 (GNU/Linux)
Comment: Using GnuPG with undefined - http://www.enigmail.net/


More information about the CITE-Forum mailing list