Sorry, I will not be at next week’s meeting to discuss. Feedback from Angelos today:
With the help from Gobe, we have identified the standards where GeoServer can be a Reference Implementation (and get certified for free) under the OGC/OSGeo MOU.
Yes there are several thousand collected from prior code sprints that were earmarked for this activity. So GeoServer PSC can pay this amount if we make a motion to do so.
For the Reference Implementation, can we meet the requirements? I thought we needed to host GeoServer somewhere to act as a reference implementations.
Update: I can continue communication while you are away.
Hi Peter,
thanks for following up. That’s a real bummer. 900$ twice a year seems too much from my point of view.
Running CITE is a benefit in itself, provides good test coverage and allows us to try deeper changes
without the fear of breaking compatibility.
At the very least, I would skip certifying GeoPackage, what we submitted is just one file after all… the test
is nowhere near comprehensive and not everyone is using that output format.
While it’s just 150$ saving, in the long run it’s going to be 300$ per year that we can dedicate to other
activities.
For the others, I’m undecided… we might skip WMS 1.1 and WFS 1.1 too, and just certify the latest?
Also consider we have two more CITE tests in the pipe for the next round, OGC API Tiles and
KML, with OGC API Processes that might join the group as well between here and September.
Running the GML suites for simple and complex features could also be of interest, in time.
In the long run, I’d take a stance of certifying what’s important, and leave “implementing” with
a link to our successful test results for less important bits that we’d have to pay for
(e.g., download the suite results and make them available maybe?)
I am rejoining this discussion on the value of certification.
The CITE certification seems to be the “list cost” for community at $150. I do not see anything specific to OSGeo MOU (based prices listed here ).
To reduce the price requires OSGeo to negotiate, or to cover via standards committee? @kalxas does the OSGeo Standards committee in general have budget (or is in position to make a budget request) for project certifications.
I like the idea of certifying 2.27.0 release for what we can (at a cost of $900).
If and then switching to an more minimum costs, focused certifying only for the latest of each standard: $600
GeoPackage 1.2 Certified
GeoTiff 1.1 Reference Implementation
OGC API - Features 1.0 Certified
WCS 2.0.1, WCS 1.1.1 and WCS 1.0 Reference Implementation
I know we could reduce that $300 more by forgoing GeoPackage and GeoTIFF. While the tests provide minimal value, the assurance I think provides some value for users of GeoServer.
Hi Jody,
there is a bit of confusion but I agree on the general statement, which seems to be let’s “splurge” for the first round and be more conservative in the long run.
The list of services for which we have to pay is the following mind:
GeoPackage 1.2
WMS 1.1.1
WMS 1.3.0
OGC API - Features 1.0
WFS 1.1.0
WFS 2.0
For GeoTiff we can be reference implementation and not pay.
I agree on the idea of paying on the latest for important services: WFS, WMS, Features.
GeoPackage wise… it’s mostly going to be for shows.
The GeoPackage test can inspect only a single GeoPackage, and we can certify based on a single run.
So the meaning of the certification will be only one of the following (I don’t know which one Peter used):
Single table vector gpkg (topp:states, 4326)
Multiple-tables vector gpkg (ne: countries and ne:disputed areas, 4326)
Single table vector gpkg, with project CRS (sf:archsites)
Single table raster gpkg (ne:countries png map tiles)
Picking just one of them is rather underwhelming, it’s a “sticker certication”.
(and mind, GeoTIFF is the same unfortunately, but at least we won’t pay for it).
For a comprehensive test in a single GeoPackage we’d have to use the gs-gpkg community module, I believe the WPS process in it can do multiple tables in various forms… however certifying based on something that’s not even supported is not great IMHO.
I’d favor an approach were we pay for what’s both important (user wise) and substantial (test wise) and then link to a document where we explain what we test, how we test, and the rationale behind the “implementing” bits.