Hi,
TL;DR optional GeoSPARQL support is so dirty to implement, so I want to
discuss what to do.
as the last part to have GeoSPARQL in KiWi (MARMOTTA-584[1][2]) merged into
develop, in the last days I've been approaching (fighting) to make the need
of PostGIS option {MARMOTTA-638 [3]) and try to finally work on a new
release with this feature.
As discussed [4], the advantage of getting MARMOTTA-638 done would be to
lower the update requirements (i.e., if you don't want GeoSPARQL you would
not need to have PostGIS extension installed in PostgreSQL).
The disadvantage is that complicates quite some the implementation: as we
extensively use prepare statements you would need to have two db schemas
(one with a fake geometry, another one with an actual geometry), which
complicates everything. Locally I have a path I'm very unsatisfied with its
quality, and I'm pretty sure will be the source of many issues in the near
future. Therefore I; m not completely sure I want to continue that path,
but I think it's an aspect that requires further discussion.
So, summarizing, we have two options:
1) Spend some more time (hard to estimate) to get GeoSPARQL optional
2) Forget
My vote now goes to the second option, because: a) it make the source base
far more maintainable; b) after all PostGIS is widely supported and trivial
to install; and c) this feature is blocking 3.4.0 for too long. But I'd
like to listen to the opinion of the community to really decide what to do.
What do you think, guys?
Thanks for your time.
Cheers,
[1] https://issues.apache.org/jira/browse/MARMOTTA-584
[2] https://github.com/apache/marmotta/tree/MARMOTTA-584
[3] https://issues.apache.org/jira/browse/MARMOTTA-638
[4] http://markmail.org/message/3u55tk4xe4g2qgoa
--
Sergio Fernández
Partner Technology Manager
Redlink GmbH
m: +43 6602747925
e: [email protected]
w: http://redlink.co
