Am 02.08.2012 12:53, schrieb Melanie Reiplinger:
Am 02.08.2012 12:34, schrieb Alessandro Adamou:
On 8/2/12 12:17 PM, Melanie Reiplinger wrote:
I just noticed that the same problem (303 at POST, then automatical
GET -> time out) exists for the session manager (when loading an
ontology into a session). This is a bit strange, since I'm pretty
sure that my tests for the session manager all worked out before my
last updates of Stanbol.
Had you tested using the browser or curl?
browser.
Also, when you rebuilt Stanbol did you remember to replace the
Clerezza - SCB Jena TDB Storage Provider with the one from the
0.6-incubating-SNAPSHOT? Your delays could be due to many databases
being created by the old implementation.
o, sorry, I did not check for that since simple updating of components
didn't make any change here. But my last rebuild was a complete
re-installation, that's why now I sitll have the 0.5 Version there.
Thank you. Might be this explains it all, I will test it and then report.
Didn't help very much. First it looked like with the single dataset tc
provider it worked better. Still, the GET requests failed (I see that in
the network log of the browser), and still, they took several seconds,
but at least they didn't cause a time out. But after a few minutes of
testing it was just the same as before: time out and failure of
uploading for ontologies.
btw: is it possible meanwhile to delete a scope from a session by
curl -X DELETE .../ontonet/session?scopeid=<scopeID>
? I'm just asking; this is no urgent issue.
(@Rupert is it enough to just build and install the
rdf.jena.tdb.storage from the trunk to make sure it uses the single
dataset tc provider?)
But then you said that the plain GET behaves now just like it has
before (showing the ontology axioms).
Indeed, it still shows the axioms. I will change the redirect at some
point.
Alessandro
Am 01.08.2012 12:04, schrieb Melanie Reiplinger:
Am 01.08.2012 12:01, schrieb Alessandro Adamou:
On 8/1/12 11:54 AM, Melanie Reiplinger wrote:
Hi Alessandro,
the GET for the test-response you sent takes less than a second
for me. So then it is a bit strange that I'm getting these time
out problems. Do you know if there's a way to avoid the
subsequent GET after uploading the ontology?
AFAIK no, as browsers react automatically on 303 responses.
If you want, while I wait to implement Ajax forms that redirect on
201, I can change the See Other Location header to the same as the
scope.
yes, maybe this is the fastest and best solution for the time
being, it would be great if you could do this. At least then this
thing stops blocking my tests.
g
melanie
Alessandro
Am 01.08.2012 11:47, schrieb Alessandro Adamou:
On 8/1/12 10:54 AM, Melanie Reiplinger wrote:
When I load an ontology upon a scope, using multipart/form-data
(i.e., FormData object in Javascript), then automatically after
this POST (if ending with 303), a GET is done by the Browser on
the URL of the ontology. I guess this is due to the
POST/Redirect/GET pattern.
Yes. Actually I would like the it to return 201 Created, but
then I'd need all the POST request in the HTML forms to be
handled by Ajax instead of standard forms. Not sure yet on the
best way to do that though (I heard there's a forms extension
for jQuery that handles Ajax forms?).
The problem now is that this subsequent GET fails as time out.
When doing the same GET via curl, it takes awfully long before
it spits out the data belonging to the ontology.
How long is it taking? When I load the OMV onto a scope (e.g.
from [1]), the subsequent GET responds in a couple of seconds on
my laptop. In general it has handled graphs with 100k triples
okay. How long is a GET taking for you on
http://bernina.cs.unibo.it/stanbol/ontonet/ontology/test-response/http://omv.ontoware.org/2005/05/ontology
?
Everything is ok if I specify -H "Accept:application/rdf+xml"
for the GET, but I guess that the automatic GET done by the
browser does a plain GET and therefore is kicked out for time
out reason.
Why does the plain GET take so long?
Because at the moment the plain GET also shows the ontology
axioms (in Manchester syntax) just like the other GETs do. Which
means it takes as much time as to export the stored graph to
that format. But again, this should hold for ontologies much
larger than the OMV...
Anyway, most likely the ontology code will disappear from the
HTML response. I'm not liking it that much and it can cause
trouble for large ontologies. I would like it to replace it with
a summary, some labels, maybe also key concepts...
Best,
Alessandro
[1] http://ontologydesignpatterns.org/ont/iks/kres/omv.owl
Am 31.07.2012 16:29, schrieb Alessandro Adamou:
Hi Melanie,
On 7/31/12 1:52 PM, Melanie Reiplinger wrote:
I did a complete re-installation and rebuild last night. Now
my bugs are gone, the delete for the contenthub works and I
can see e.g. the modified RESTful interface of the
ontologymanager (which I did not see before, in spite of
updating). The curl command for adding 2 scopes at once to
some existing session now works for me. :-)
I forgot to tell you that in the latest update you had to
rebuild and reinstall commons/owl too. However this happened
automatically when you rebuilt the whole Stanbol.
But anyway, to make sure the update takes place without
rebuilding everything, perhaps the best thing to do is to
install the single bundles as you build them, like this (while
Stanbol is up and running):
mvn clean install -PinstallBundle
-Dsling.url=http://[stanbol]/system/console
if you have changed the admin credentials to something other
than admin:admin , also add -Dsling.user=[admin-username]
-Dsling.password=[admin-password]
Small update: I finally started working on the D of CRUD for
stored ontologies! (i.e. on the /ontonet endpoint). Not sure
how long it will take but I'd like to have a primitive
implementation by the end of the week.
probably the first implementation will be conservative, i.e.
if there are any handles on the ontology you will get a 409
Conflict or 403 Forbidden, not sure yet. You will first have
to clear it of any handles from scopes or sessions.
Also for loading 2 libraries on a scope at creation, I
succeeded after I had escaped the '&' symbol in the URL. This
is a bit surprising for me, because I have used curl commands
with the '&' several times and never had to escape it until
now. I'll see how this works when sending the request via
XmlHttpRequest using CORS, maybe then I don't need to escape it.
This is a command line feature (I assume you're using curl
from a Unix shell). In Bash shells, the '&' character means
"fork this command as a background process" and everything
after it is ignored unless you escape it. You don't have to
escape it for Ajax requests, so I guess that's why it worked
for your XmlHttpRequest in JavaScript. Are you sure that in
your previous curl calls the query parameters from the second
on (i.e. that followed '&') were actually considered?
Best,
Alessandro
Am 30.07.2012 14:16, schrieb Alessandro Adamou:
Hi Melanie,
On 7/30/12 1:27 PM, Melanie Reiplinger wrote:
Sorry for my late reply, but I took a 2-week break from the
job to intensively study and revise for my exams that took
place last week.
I understand. I hope they all went fine. I too am
extensively working on my Ph.D. thesis (of which the Stanbol
ontology manager is but a partial implementation BTW).
Can I be curious? I'd like to ask you what is the use case
you're using ontonet for...
I tried to pass e.g. several scopes to load them into a
session, but I'm doing something wrong. In what form is the
list of scopes expected? I tried specifying 2 -F
parameters, like
curl -i -X POST -F scope=scope1 -F scope=scope2
http://<stanbol>/ontonet/session/mysession
Actually that is correct and it worked for me, provided that
scope1 scope2 and mysession were all created earlier. What
response are you receiving? You should just get a 303 on the
session resource, but then if you request an RDF/OWl of the
session, as in
curl -H "Accept: text/turtle"
http://<stanbol>/ontonet/session/mysession
you should see an import statement for each scope.
Perhaps some day I will make scope attachment stateless via
GET.
Trivial question: did you first svn update, rebuild, cleanup
the Stanbol installation etc?
The same for passing several corereg or coreont parameters:
How do I list them for the curl command? Something like
<stanbol>/ontology/scope1?corereg=lib1&corereg=lib2
won't work. Do I have to specify a list?
Again, that seems correct. But remember that if you are
using curl via command line you must escape the '&'
character. Example:
curl -i -X PUT
[stanbol]/ontonet/ontology/scope1?corereg=http://stanbol.apache.org/ontologies/registries/stanbol_network/Foundational\&corereg=http://stanbol.apache.org/ontologies/registries/stanbol_network/W3C
Hope this is helping
Alessandro
Am 23.07.2012 18:44, schrieb Alessandro Adamou:
Hi,
Since you had been following its development more closely
than others, I just thought you might be interested in the
latest improvements in the Stanbol ontology manager.
Please do not hesitate to try it out to see if it better
fits your needs and give whatever feedback you would like.
(NOTE I still need to update the documentation. That's my
next TODO)
All the best,
Alessandro
-------- Original Message --------
Subject: Latest Ontology Manager improvements
Date: Mon, 23 Jul 2012 18:42:18 +0200
From: Alessandro Adamou <[email protected]>
Reply-To: [email protected]
Organization: Alma Mater Studiorum - Università di
Bologna
To: [email protected]
Hi,
just wanted to share about the work I did on the Stanbol
Ontology Manager over the last couple of months, as it is
also based on requests that surfaced on the list:
NEW:
* Persistence and portability of ontology network
structures: relations between scopes, sessions and the
ontologies managed by them are now stored in a special
metadata graph. Scopes and sessions are restored as soon
as OntoNet activates (though sessions are restored in an
inactive state). I will publish the OWL schema soon.
* Configurable connectivity policies. Depending on the
chosen policy, more import statements will be declared
across scopes, spaces and session. This has shown to
greatly affect OWL axiom interpretation and reasoning
results. Default scheme is TIGHT.
* REST service for attaching/detaching multiple scopes to
a session at once. Just POST as many "scope={scopeID}"
form parameters as you like to the session resource. Any
previously attached scope not included will be detached
automatically
IMPROVED:
* Speedups. Reduced the likelihood of copying triples from
Clerezza-based input sources. This should reduce the
length of ontology load operations by about 50%.
* RESTful services for exporting ontology networks no
longer depend on the configured namespaces for OntoNet
components. URI prefixes are now extracted from the
request URL. Only the Java API depends on this
configuration now.
* HTML client can now create and delete scopes and sessions.
* POST service for creating sessions and letting Stanbol
choose their ID (will be returned as the Location header
in the 201 Created response)
* Most creation methods now return 201 Created with the
resource Location. It's still 303 for some service because
the HTML client uses POST forms and would not be able to
capture a 201 response and act thereupon.
* Ontology -> Stored Graph reconciliation policies (
STANBOL-426). It is now possible to obtain a "public key"
for each supplied OntologyID (incl. owl:versionIRI) or
physical URL.
* Parameters "corereg" and "coreont" for creating scopes
with ontologies and libraries embedded in their core
spaces can now have multiple values in a single call.
* Added backward links to the HTML interface
FIXED:
* CORS support (thanks Melanie for testing!)
* Parsing formats other than RDF/XML without specifying
the parser to try was often failing as the streams were
not being reset/recreated. This was also causing failures
if imported ontologies were in a format different than
RDF/XML.
* Graph naming policy for anonymous ontologies was causing
trouble when restoring components. Moved it out of
commons/owl
* Broke computational loops caused by cycles in import
graphs (a consequence of migrating to the Clerezza API).
* Got rid of the obsolete OWL API-based implementation
Working on a couple more things right now, then once the
OntoNet renaming takes place I would call for a new
ontologymanager release.
Best,
Alessandro
--
M.Sc. Alessandro Adamou
Alma Mater Studiorum - Università di Bologna
Department of Computer Science
Mura Anteo Zamboni 7, 40127 Bologna - Italy
Semantic Technology Laboratory (STLab)
Institute for Cognitive Science and Technology (ISTC)
National Research Council (CNR)
Via Nomentana 56, 00161 Rome - Italy
"I will give you everything, just don't demand anything."
(Ettore Petrolini, 1917)
Not sent from my iSnobTechDevice
--
M.Sc. Alessandro Adamou
Alma Mater Studiorum - Università di Bologna
Department of Computer Science
Mura Anteo Zamboni 7, 40127 Bologna - Italy
Semantic Technology Laboratory (STLab)
Institute for Cognitive Science and Technology (ISTC)
National Research Council (CNR)
Via Nomentana 56, 00161 Rome - Italy
"I will give you everything, just don't demand anything."
(Ettore Petrolini, 1917)
Not sent from my iSnobTechDevice
--
M.Sc. Alessandro Adamou
Alma Mater Studiorum - Università di Bologna
Department of Computer Science
Mura Anteo Zamboni 7, 40127 Bologna - Italy
Semantic Technology Laboratory (STLab)
Institute for Cognitive Science and Technology (ISTC)
National Research Council (CNR)
Via Nomentana 56, 00161 Rome - Italy
"I will give you everything, just don't demand anything."
(Ettore Petrolini, 1917)
Not sent from my iSnobTechDevice
--
M.Sc. Alessandro Adamou
Alma Mater Studiorum - Università di Bologna
Department of Computer Science
Mura Anteo Zamboni 7, 40127 Bologna - Italy
Semantic Technology Laboratory (STLab)
Institute for Cognitive Science and Technology (ISTC)
National Research Council (CNR)
Via Nomentana 56, 00161 Rome - Italy
"I will give you everything, just don't demand anything."
(Ettore Petrolini, 1917)
Not sent from my iSnobTechDevice
