Re: OWL2 serialized as JSON?

From: Bob Ferris <zazi@...>
Subject: Re: OWL2 serialized as JSON?
Date: Wed, 6 Apr 2011 03:47:20 -0500

> Hi Chris,
> 
> On 4/5/2011 11:29 PM, Chris Mungall wrote:
>> Is there a de-facto standard way of serializing OWL2 as JSON? I'm aware of RDF-over-JSON efforts, but
something more OWL-centric would suit my purposes better.
>>
>> I recall a lightning talk at OWLED2007 that showed something like Manchester Syntax in JSON, but I don't
know if this idea has advanced further. I think ideally there might be a frame-style modeled after MS (but
including GCIs), and an axiom-style modeled after the functional syntax. It seems the most predictable
way to do the latter would be to have a single object per axiom, and to use the non-terminals on the RHS of the
production rules as names in the name-value pairs.
>>
>> I'm hoping someone has already provided a specification - and/or an OWLAPI implementation?
> 
> I'm not aware of any OWL2/JSON serialization. However, I'm wondering 
> whether this is really necessary, since OWL can be represented via the 
> knowledge representation structure RDF Model quite well. Today there are 
> multiple proposals for RDF/JSON serialization (see [1]) available and 
> the new RDF WG is working on a standard recommendation. Maybe you should 
> wait for this.
> What are the benefits of having a separate OWL2/JSON serialization format?
> 
> Cheers,
> Bob
> 
> [1] http://www.w3.org/2011/rdf-wg/wiki/JSON-Serialization-Examples
(Continue reading)

Bob Ferris | 6 Apr 13:46 2011
Picon

Re: OWL2 serialized as JSON?

Hi Peter,

On 4/6/2011 1:18 PM, Peter Frederick Patel-Schneider wrote:
> From: Bob Ferris<zazi@...>
> Subject: Re: OWL2 serialized as JSON?
> Date: Wed, 6 Apr 2011 03:47:20 -0500
>
>> Hi Chris,
>>
>> On 4/5/2011 11:29 PM, Chris Mungall wrote:
>>> Is there a de-facto standard way of serializing OWL2 as JSON? I'm aware of RDF-over-JSON efforts, but
something more OWL-centric would suit my purposes better.
>>>
>>> I recall a lightning talk at OWLED2007 that showed something like Manchester Syntax in JSON, but I don't
know if this idea has advanced further. I think ideally there might be a frame-style modeled after MS (but
including GCIs), and an axiom-style modeled after the functional syntax. It seems the most predictable
way to do the latter would be to have a single object per axiom, and to use the non-terminals on the RHS of the
production rules as names in the name-value pairs.
>>>
>>> I'm hoping someone has already provided a specification - and/or an OWLAPI implementation?
>>
>> I'm not aware of any OWL2/JSON serialization. However, I'm wondering
>> whether this is really necessary, since OWL can be represented via the
>> knowledge representation structure RDF Model quite well. Today there are
>> multiple proposals for RDF/JSON serialization (see [1]) available and
>> the new RDF WG is working on a standard recommendation. Maybe you should
>> wait for this.
>> What are the benefits of having a separate OWL2/JSON serialization format?
>>
>> Cheers,
(Continue reading)

Re: OWL2 serialized as JSON?

From: Bob Ferris <zazi@...>
Subject: Re: OWL2 serialized as JSON?
Date: Wed, 6 Apr 2011 06:46:30 -0500

> Hi Peter,
> 
> On 4/6/2011 1:18 PM, Peter Frederick Patel-Schneider wrote:
>> From: Bob Ferris<zazi@...>
>> Subject: Re: OWL2 serialized as JSON?
>> Date: Wed, 6 Apr 2011 03:47:20 -0500
>>
>>> Hi Chris,
>>>
>>> On 4/5/2011 11:29 PM, Chris Mungall wrote:
>>>> Is there a de-facto standard way of serializing OWL2 as JSON? I'm aware of RDF-over-JSON efforts, but
something more OWL-centric would suit my purposes better.
>>>>
>>>> I recall a lightning talk at OWLED2007 that showed something like Manchester Syntax in JSON, but I
don't know if this idea has advanced further. I think ideally there might be a frame-style modeled after MS
(but including GCIs), and an axiom-style modeled after the functional syntax. It seems the most
predictable way to do the latter would be to have a single object per axiom, and to use the non-terminals on
the RHS of the production rules as names in the name-value pairs.
>>>>
>>>> I'm hoping someone has already provided a specification - and/or an OWLAPI implementation?
>>>
>>> I'm not aware of any OWL2/JSON serialization. However, I'm wondering
>>> whether this is really necessary, since OWL can be represented via the
>>> knowledge representation structure RDF Model quite well. Today there are
>>> multiple proposals for RDF/JSON serialization (see [1]) available and
>>> the new RDF WG is working on a standard recommendation. Maybe you should
(Continue reading)

Phillip Lord | 6 Apr 17:07 2011
Picon
Picon

Re: OWL2 serialized as JSON?


Peter Frederick Patel-Schneider <pfps@...> writes:
> I would argue instead that RDF/JSON for OWL axioms would be much more
> confusing than a direct transformation.

I think that Chris, who asked for this in the first place, has already
agreed with this position. For what it's worth, I would also agree. If I
want to operate in JSON, then I don't really want to have to think about
the RDF syntactic representation of OWL at the same time. 

As a reduction to absurdity, for example, one way to get a JSON
representation of OWL, would be to parse the XML of the OWL into DOM,
then JSONify the DOM objects. Or, even more trivially, take the XML, add
a quote to the start, a quote to the end, and escape any strings in the
middle, and we something which is a valid JSON object. 

Both of these sort of miss the point, which is presumably to get a
representation of the data in the OWL stuffed into a data model which
looks like OWL. I agree this comes with the additional costs of having
an additional data model, but then the same argument holds for RDF and,
for example, XML or even Unicode. 

Phil

Nathan | 10 Apr 09:51 2011

Re: OWL2 serialized as JSON?

Peter Frederick Patel-Schneider wrote:
> From: Bob Ferris <zazi@...>
> Subject: Re: OWL2 serialized as JSON?
> Date: Wed, 6 Apr 2011 06:46:30 -0500
> 
>> Hi Peter,
>>
>> On 4/6/2011 1:18 PM, Peter Frederick Patel-Schneider wrote:
>>> From: Bob Ferris<zazi@...>
>>> Subject: Re: OWL2 serialized as JSON?
>>> Date: Wed, 6 Apr 2011 03:47:20 -0500
>>>
>>>> Hi Chris,
>>>>
>>>> On 4/5/2011 11:29 PM, Chris Mungall wrote:
>>>>> Is there a de-facto standard way of serializing OWL2 as JSON? I'm aware of RDF-over-JSON efforts, but
something more OWL-centric would suit my purposes better.
>>>>>
>>>>> I recall a lightning talk at OWLED2007 that showed something like Manchester Syntax in JSON, but I
don't know if this idea has advanced further. I think ideally there might be a frame-style modeled after MS
(but including GCIs), and an axiom-style modeled after the functional syntax. It seems the most
predictable way to do the latter would be to have a single object per axiom, and to use the non-terminals on
the RHS of the production rules as names in the name-value pairs.
>>>>>
>>>>> I'm hoping someone has already provided a specification - and/or an OWLAPI implementation?
>>>> I'm not aware of any OWL2/JSON serialization. However, I'm wondering
>>>> whether this is really necessary, since OWL can be represented via the
>>>> knowledge representation structure RDF Model quite well. Today there are
>>>> multiple proposals for RDF/JSON serialization (see [1]) available and
>>>> the new RDF WG is working on a standard recommendation. Maybe you should
(Continue reading)

Eric Prud'hommeaux | 10 Apr 16:28 2011
Picon

Re: OWL2 serialized as JSON?

* Nathan <nathan@...> [2011-04-10 08:51+0100]
> Peter Frederick Patel-Schneider wrote:
> >From: Bob Ferris <zazi@...>
> >Subject: Re: OWL2 serialized as JSON?
> >Date: Wed, 6 Apr 2011 06:46:30 -0500
> >
> >>Hi Peter,
> >>
> >>On 4/6/2011 1:18 PM, Peter Frederick Patel-Schneider wrote:
> >>>From: Bob Ferris<zazi@...>
> >>>Subject: Re: OWL2 serialized as JSON?
> >>>Date: Wed, 6 Apr 2011 03:47:20 -0500
> >>>
> >>>>Hi Chris,
> >>>>
> >>>>On 4/5/2011 11:29 PM, Chris Mungall wrote:
> >>>>>Is there a de-facto standard way of serializing OWL2 as JSON? I'm aware of RDF-over-JSON efforts,
but something more OWL-centric would suit my purposes better.
> >>>>>
> >>>>>I recall a lightning talk at OWLED2007 that showed something like Manchester Syntax in JSON, but I
don't know if this idea has advanced further. I think ideally there might be a frame-style modeled after MS
(but including GCIs), and an axiom-style modeled after the functional syntax. It seems the most
predictable way to do the latter would be to have a single object per axiom, and to use the non-terminals on
the RHS of the production rules as names in the name-value pairs.
> >>>>>
> >>>>>I'm hoping someone has already provided a specification - and/or an OWLAPI implementation?
> >>>>I'm not aware of any OWL2/JSON serialization. However, I'm wondering
> >>>>whether this is really necessary, since OWL can be represented via the
> >>>>knowledge representation structure RDF Model quite well. Today there are
> >>>>multiple proposals for RDF/JSON serialization (see [1]) available and
(Continue reading)

Bijan Parsia | 6 Apr 14:00 2011
Picon
Picon

Re: OWL2 serialized as JSON?

On 6 Apr 2011, at 12:46, Bob Ferris wrote:

> Hi Peter,
[snip]
> However, then you probably confuse people even more who are trying to get into Semantic Web, or?

I think the likelihood is small, esp. that there'd be a significant marginal degradation.

We don't and shouldn't optimize everything for the sake of newbies (though obviously we shouldn't be hostile).

> If we would have two separate serializations formats, then you have to teach people RDF/JSON and OWL/JSON,

I imagine that the people who need to know both is rather small.

The main point is that there's a requirement for a concrete serialization in JSON of something close to the
OWL abstract syntax. This is not (easily) met by any RDF/JSON serialization.

It would behoove you to understand the requirement :) You may think meeting it isn't, globally, a good idea,
but you should understand what you aren't meeting.

> and convience them from their benefits and existence.

I don't see why we have to do that for arbitrary people. (Existence is easy :)).

> The power of RDF Model is that it is a knowledge representation structure for the vocabulary level and the
instantitation level.

Er...that's not really a power.

> Otherwise, you would (prefer to) use OWL/JSON for vocabulary level serializations and RDF/JSON for
(Continue reading)

Chris Mungall | 6 Apr 23:27 2011

Re: OWL2 serialized as JSON?

On Apr 6, 2011, at 4:46 AM, Bob Ferris wrote:

> Hi Peter,
> 
> On 4/6/2011 1:18 PM, Peter Frederick Patel-Schneider wrote:
>> From: Bob Ferris<zazi@...>
>> Subject: Re: OWL2 serialized as JSON?
>> Date: Wed, 6 Apr 2011 03:47:20 -0500
>> 
>>> Hi Chris,
>>> 
>>> On 4/5/2011 11:29 PM, Chris Mungall wrote:
>>>> Is there a de-facto standard way of serializing OWL2 as JSON? I'm aware of RDF-over-JSON efforts, but
something more OWL-centric would suit my purposes better.
>>>> 
>>>> I recall a lightning talk at OWLED2007 that showed something like Manchester Syntax in JSON, but I
don't know if this idea has advanced further. I think ideally there might be a frame-style modeled after MS
(but including GCIs), and an axiom-style modeled after the functional syntax. It seems the most
predictable way to do the latter would be to have a single object per axiom, and to use the non-terminals on
the RHS of the production rules as names in the name-value pairs.
>>>> 
>>>> I'm hoping someone has already provided a specification - and/or an OWLAPI implementation?
>>> 
>>> I'm not aware of any OWL2/JSON serialization. However, I'm wondering
>>> whether this is really necessary, since OWL can be represented via the
>>> knowledge representation structure RDF Model quite well. Today there are
>>> multiple proposals for RDF/JSON serialization (see [1]) available and
>>> the new RDF WG is working on a standard recommendation. Maybe you should
>>> wait for this.
>>> What are the benefits of having a separate OWL2/JSON serialization format?
(Continue reading)

Jerven Bolleman | 7 Apr 09:33 2011
Picon

Re: OWL2 serialized as JSON?

Hi Chris, All,

I have the feeling that you are going about this the wrong way round.
I would first write a compelling JS api to deal with OWL concepts. And 
later if necessary design an optimized serialization format.

This will be better because
a) if you import someone elses OWL you can accept multiple 
serializations. As the serialization to JS object model is independent 
of the final API.
b) it makes it easier for developers like me to try it out. No need to 
change server side code.

If you use serialization as your API.
a) You will need convert imported foreign OWL server side. Increasing 
your load and maintenance problems in the future.
b) Potential users need to cross one more bridge before the can try your 
API out.
c) Serialization as an API introduces hard limits to re-usability and 
extendability.

Of course I understand if you need to meet your own needs first.

Regards,
Jerven

On 04/06/2011 11:27 PM, Chris Mungall wrote:
> On Apr 6, 2011, at 4:46 AM, Bob Ferris wrote:
>
>> Hi Peter,
(Continue reading)

Bijan Parsia | 7 Apr 14:54 2011
Picon
Picon

Re: OWL2 serialized as JSON?

On 7 Apr 2011, at 08:33, Jerven Bolleman wrote:

> Hi Chris, All,
> 
> I have the feeling that you are going about this the wrong way round.
> I would first write a compelling JS api to deal with OWL concepts. And later if necessary design an
optimized serialization format.

Actually this is pretty close to what I proposed to do.

The structure spec defines a quite nice API for OWL ontologies:
	http://www.w3.org/TR/2009/REC-owl2-syntax-20091027/
(The Manchester OWL API adheres to this.)

The XML Serialization mirrors this closely:
	http://www.w3.org/TR/2009/REC-owl2-xml-serialization-20091027/

All other serializations (Manchester Syntax, RDF syntax) have a mapping to the abstract model.

Although there are some issues with things for serialization (e.g., prefixes). I'll try to separate these
out (as I'm currently doing for XML).

Thus, the idea is to produce something close to this (with perhaps a few tweaks) so that, e.g., the
structural spec serves as documentation for the API.

I would generally recommend this as the preferred way to handle additional mappings and concrete formats.
That was certainly the intent of the design.

Cheers,
Bijan.
(Continue reading)

Nathan | 24 Apr 01:28 2011

Re: OWL2 serialized as JSON?

Bijan Parsia wrote:
> On 7 Apr 2011, at 08:33, Jerven Bolleman wrote:
> 
>> Hi Chris, All,
>>
>> I have the feeling that you are going about this the wrong way round.
>> I would first write a compelling JS api to deal with OWL concepts. And later if necessary design an
optimized serialization format.
> 
> Actually this is pretty close to what I proposed to do.
> 
> The structure spec defines a quite nice API for OWL ontologies:
> 	http://www.w3.org/TR/2009/REC-owl2-syntax-20091027/
> (The Manchester OWL API adheres to this.)
> 
> The XML Serialization mirrors this closely:
> 	http://www.w3.org/TR/2009/REC-owl2-xml-serialization-20091027/
> 
> All other serializations (Manchester Syntax, RDF syntax) have a mapping to the abstract model.
> 
> Although there are some issues with things for serialization (e.g., prefixes). I'll try to separate
these out (as I'm currently doing for XML).
> 
> Thus, the idea is to produce something close to this (with perhaps a few tweaks) so that, e.g., the
structural spec serves as documentation for the API.
> 
> I would generally recommend this as the preferred way to handle additional mappings and concrete
formats. That was certainly the intent of the design.

So, things I can see on the cards:
(Continue reading)

Pat Hayes | 24 Apr 04:08 2011
Picon

Re: OWL2 serialized as JSON?


On Apr 23, 2011, at 6:28 PM, Nathan wrote:

> Bijan Parsia wrote:
>> On 7 Apr 2011, at 08:33, Jerven Bolleman wrote:
>>> Hi Chris, All,
>>> 
>>> I have the feeling that you are going about this the wrong way round.
>>> I would first write a compelling JS api to deal with OWL concepts. And later if necessary design an
optimized serialization format.
>> Actually this is pretty close to what I proposed to do.
>> The structure spec defines a quite nice API for OWL ontologies:
>> 	http://www.w3.org/TR/2009/REC-owl2-syntax-20091027/
>> (The Manchester OWL API adheres to this.)
>> The XML Serialization mirrors this closely:
>> 	http://www.w3.org/TR/2009/REC-owl2-xml-serialization-20091027/
>> All other serializations (Manchester Syntax, RDF syntax) have a mapping to the abstract model.
>> Although there are some issues with things for serialization (e.g., prefixes). I'll try to separate
these out (as I'm currently doing for XML).
>> Thus, the idea is to produce something close to this (with perhaps a few tweaks) so that, e.g., the
structural spec serves as documentation for the API.
>> I would generally recommend this as the preferred way to handle additional mappings and concrete
formats. That was certainly the intent of the design.
> 
> So, things I can see on the cards:
> 
> - JSON serialization of OWL 2
> - WebIDL API (targeting Javascript and Java)
> - OWL 2 (perhaps a subset of) merging with JSON-Schema
> - RDF (perhaps a subset of) merging with JSON
(Continue reading)

Chris Mungall | 1 Dec 19:17 2011

Re: OWL2 serialized as JSON?


On Apr 7, 2011, at 5:54 AM, Bijan Parsia wrote:

> On 7 Apr 2011, at 08:33, Jerven Bolleman wrote:
> 
>> Hi Chris, All,
>> 
>> I have the feeling that you are going about this the wrong way round.
>> I would first write a compelling JS api to deal with OWL concepts. And later if necessary design an
optimized serialization format.
> 
> Actually this is pretty close to what I proposed to do.
> 
> The structure spec defines a quite nice API for OWL ontologies:
> 	http://www.w3.org/TR/2009/REC-owl2-syntax-20091027/
> (The Manchester OWL API adheres to this.)
> 
> The XML Serialization mirrors this closely:
> 	http://www.w3.org/TR/2009/REC-owl2-xml-serialization-20091027/
> 
> All other serializations (Manchester Syntax, RDF syntax) have a mapping to the abstract model.
> 
> Although there are some issues with things for serialization (e.g., prefixes). I'll try to separate
these out (as I'm currently doing for XML).
> 
> Thus, the idea is to produce something close to this (with perhaps a few tweaks) so that, e.g., the
structural spec serves as documentation for the API.
> 
> I would generally recommend this as the preferred way to handle additional mappings and concrete
formats. That was certainly the intent of the design.
(Continue reading)

Chris Mungall | 7 Aug 00:45 2013

Re: OWL2 serialized as JSON?

Remember this thread?

It stirred a bit of discussion regarding the relative merits of a direct serialization of OWL2 into JSON vs indirect via RDF. Probably somewhat academic, as here we are some time later and there don't seem to be many people publicly shunting around OWL as JSON. I have a translation I have been using for internal purposes but would like to abandon it in favor of something more standard.

I have shifted somewhat in the direction of an RDF-oriented solution. IMany of the OWL class axioms I work with tend to generate fairly verbose RDF (and consequently JSON derived from this). However, it's likely that *any* translation to JSON will likely be ugly for my axioms.

It seems JSON-LD has been gaining traction, and has nice features for avoid verbosity. Is there any move to have a standard <at> context (perhaps served from a standard URL) for OWL 2? Rather than having an abstract discussion about relative merits it might help to see some concrete examples of ontologies of varying levels of complexity translated to JSON and compacted as JSON-LD. I'm particularly interested in any JSON-LD tricks could be used for a more compact encoding of axiom annotations.



On Thu, Dec 1, 2011 at 10:17 AM, Chris Mungall <cjmungall-/3juihCSby0@public.gmane.org> wrote:

On Apr 7, 2011, at 5:54 AM, Bijan Parsia wrote:

> On 7 Apr 2011, at 08:33, Jerven Bolleman wrote:
>
>> Hi Chris, All,
>>
>> I have the feeling that you are going about this the wrong way round.
>> I would first write a compelling JS api to deal with OWL concepts. And later if necessary design an optimized serialization format.
>
> Actually this is pretty close to what I proposed to do.
>
> The structure spec defines a quite nice API for OWL ontologies:
>       http://www.w3.org/TR/2009/REC-owl2-syntax-20091027/
> (The Manchester OWL API adheres to this.)
>
> The XML Serialization mirrors this closely:
>       http://www.w3.org/TR/2009/REC-owl2-xml-serialization-20091027/
>
> All other serializations (Manchester Syntax, RDF syntax) have a mapping to the abstract model.
>
> Although there are some issues with things for serialization (e.g., prefixes). I'll try to separate these out (as I'm currently doing for XML).
>
> Thus, the idea is to produce something close to this (with perhaps a few tweaks) so that, e.g., the structural spec serves as documentation for the API.
>
> I would generally recommend this as the preferred way to handle additional mappings and concrete formats. That was certainly the intent of the design.

Hi Bijan. How is this progressing?

I've written some code on top of the OWL API that generates json from either expressions or axioms. The resulting json is fairly generic and loosely corresponds to OWL-XML. Anything that is not a URI or a literal is translated to a hash with a "type" key that maps to the axiom or expression type, and an "args" arr. This is mostly for internal purposes write now - I'd like to adopt whatever de facto standard there is out there.


> Cheers,
> Bijan.
>


Holger Knublauch | 7 Apr 01:25 2011

Re: OWL2 serialized as JSON?


On Apr 6, 2011, at 9:46 PM, Bob Ferris wrote:
> However, then you probably confuse people even more who are trying to get into Semantic Web, or? If we would
have two separate serializations formats, then you have to teach people RDF/JSON and OWL/JSON, and
convience them from their benefits and existence. The power of RDF Model is that it is a knowledge
representation structure for the vocabulary level and the instantitation level. Otherwise, you would
(prefer to) use OWL/JSON for vocabulary level serializations and RDF/JSON for instatiation level
serializations. Finally, the size reduction would be a consequence of a more complex grammar, which
might be a disadvantage.

Bob,

I actually agree with your point, but you will have noticed that many people in the OWL (2) community really
don't care much about the Web and RDF aspects of the Semantic Web. So with the usual folks wanting yet
another serialization, let them drift away further and further, into complete practical irrelevance.
It's a shame that this is still done under the umbrella of the Semantic Web, but this is IMHO due to a bug in the
W3C processes: any small community of enthusiasts can work on new standards, and other people typically
don't care enough to prevent this from progressing. As soon as OWL 2 was moving away from its RDF foundation
it should have been properly decoupled, and we wouldn't even ask those questions anymore.

Sorry for the sarcasm.

Regards,
Holger

Bijan Parsia | 7 Apr 01:40 2011
Picon
Picon

Re: OWL2 serialized as JSON?

On 7 Apr 2011, at 00:25, Holger Knublauch wrote:
[snip]
> Bob,
> 
> I actually agree with your point, but you will have noticed that many people in the OWL (2) community really
don't care much about the Web

This is tendencious at best. JSON is as web as anything and being interested in it is not remotely evidence of
not caring much about the Web.

> and RDF aspects of the Semantic Web.

It's also perfectly possible to care about RDF and not think that the RDF serialization of OWL is the right
thing in every circumstance.

[snip]
> Sorry for the sarcasm.

Where was the sarcasm? I just saw a bit of ungrounded psychologizing and rather tired bad mouthing. I'm not
sure why you felt it necessary to rant this way, but hey, to each their own.

It'd be nice to get back to technical issues.

Cheers,
Bijan.

Holger Knublauch | 7 Apr 01:54 2011

Re: OWL2 serialized as JSON?

On Apr 7, 2011, at 9:40 AM, Bijan Parsia wrote:
> It'd be nice to get back to technical issues.

There are some very clear technical advantages of trying to stay compatible with RDF. For example uniform
storage and APIs via RDF triple stores. Or the extensibility of RDF-based language that has been thrown
away by hard-coding only a pre-selected choice of OWL constructs [1]. Or having a proper stack of
languages, where one language of a family can be used in conjunction with another. The two of us had already
discussed that using SPARQL and OWL together is very difficult with OWL 2-based APIs.

While I can understand the desire to have optimized solutions for your side of the story, I have seen enough
practical evidence (e.g. from our customer base) that the confusion caused by all those dialects and
serializations is actually an obstacle to the adoption of this technology.

And regarding JSON or not, of course I am all in favor of using JSON with Semantic Web languages. The new RDF
working group is already working on grammars for this. But since the OWL 2 people have departed from RDF,
they will of course now also need to duplicate this work, and again come up with something that throws away
some very fundamental ideas of the Semantic Web.

Holger

[1] http://composing-the-semantic-web.blogspot.com/2010/04/where-owl-fails.html


Gmane