You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@stanbol.apache.org by Reto Bachmann-Gmür <re...@apache.org> on 2013/05/21 18:01:44 UTC
Getting rid this build-time downloads
I run once again in a (certainly temporary) build-download exception.
(Failed to copy
https://github.com/utcompling/OpenNLP-Models/raw/58ef0c60031403e66e47ae35edaf58d3478b67af/models/es/opennlp-es-maxent-pos-es.bin)
I think we should discus on ways to making our build so that it runs
offline (given that the necessary maven dependencies are in the local repo).
- For not too big dependencies (like the one failing here) they could go to
maven central.
- Bigger data-files (not sure how big) they should be downloaded at
runtime. I think this would be be implemented with a centralized service
offering monitoring which is called by the components that need such a
data-file
I know adding new features is more fun but I think some cleaning too have a
smooth and reliable build would be an important improvement.
WDYT?
Reto
Re: Getting rid this build-time downloads
Posted by Fabian Christ <ch...@googlemail.com>.
Hi,
I agree that we should get rid of those downloads. The
DataFileProvider seems to be the right tool. Would be nice to have
some UI to make it easy for administrators to select which models to
download.
For the unit tests, I am not sure how complex it would be to create
some test models that can be distributed without any license issues.
Those models need not to be perfect - they are just used to test
Stanbol's functionality. We do not test the quality of the results,
right?
Best,
- Fabian
2013/5/22 Rupert Westenthaler <ru...@gmail.com>:
> Hi Reto,
>
> Have you seen the DataFileProvider tab in the OSGI console? This one
> could be improved to allow users to manually trigger a download for
> resources that do have the "Download-Location" key, We could also add
> an option to automatically download resources from specific servers.
>
> With that feature in place the data bundles would no longer require to
> include the data files (as Stanbol could download those at runtime).
> However we would still need some of the data files for unit testing
> (e.g. the english OpenNLP models and also the default DBpedia dataset
> are used by several unit tests).
>
> best
> Rupert
>
> On Tue, May 21, 2013 at 6:01 PM, Reto Bachmann-Gmür <re...@apache.org> wrote:
>> I run once again in a (certainly temporary) build-download exception.
>>
>> (Failed to copy
>> https://github.com/utcompling/OpenNLP-Models/raw/58ef0c60031403e66e47ae35edaf58d3478b67af/models/es/opennlp-es-maxent-pos-es.bin)
>>
>>
>> I think we should discus on ways to making our build so that it runs
>> offline (given that the necessary maven dependencies are in the local repo).
>>
>> - For not too big dependencies (like the one failing here) they could go to
>> maven central.
>> - Bigger data-files (not sure how big) they should be downloaded at
>> runtime. I think this would be be implemented with a centralized service
>> offering monitoring which is called by the components that need such a
>> data-file
>>
>> I know adding new features is more fun but I think some cleaning too have a
>> smooth and reliable build would be an important improvement.
>>
>> WDYT?
>> Reto
>
>
>
> --
> | Rupert Westenthaler rupert.westenthaler@gmail.com
> | Bodenlehenstraße 11 ++43-699-11108907
> | A-5500 Bischofshofen
--
Fabian
http://twitter.com/fctwitt
Re: Getting rid this build-time downloads
Posted by Rupert Westenthaler <ru...@gmail.com>.
Hi Reto,
Have you seen the DataFileProvider tab in the OSGI console? This one
could be improved to allow users to manually trigger a download for
resources that do have the "Download-Location" key, We could also add
an option to automatically download resources from specific servers.
With that feature in place the data bundles would no longer require to
include the data files (as Stanbol could download those at runtime).
However we would still need some of the data files for unit testing
(e.g. the english OpenNLP models and also the default DBpedia dataset
are used by several unit tests).
best
Rupert
On Tue, May 21, 2013 at 6:01 PM, Reto Bachmann-Gmür <re...@apache.org> wrote:
> I run once again in a (certainly temporary) build-download exception.
>
> (Failed to copy
> https://github.com/utcompling/OpenNLP-Models/raw/58ef0c60031403e66e47ae35edaf58d3478b67af/models/es/opennlp-es-maxent-pos-es.bin)
>
>
> I think we should discus on ways to making our build so that it runs
> offline (given that the necessary maven dependencies are in the local repo).
>
> - For not too big dependencies (like the one failing here) they could go to
> maven central.
> - Bigger data-files (not sure how big) they should be downloaded at
> runtime. I think this would be be implemented with a centralized service
> offering monitoring which is called by the components that need such a
> data-file
>
> I know adding new features is more fun but I think some cleaning too have a
> smooth and reliable build would be an important improvement.
>
> WDYT?
> Reto
--
| Rupert Westenthaler rupert.westenthaler@gmail.com
| Bodenlehenstraße 11 ++43-699-11108907
| A-5500 Bischofshofen