[PROJ] EPSG release once more busted

[Martin Desruisseaux]

Le 21/07/2026 à 21:43, Even Rouault via PROJ a écrit :

> I'm trying to integrate the latest EPSG 12.059a release and there are 
> *again* consistency errors. Not necessary big deals, usually I deal 
> with that directly with them in a diplomatic way, but I'm fed up of 
> that. What kind of pressure can we apply on EPSG so they implement 
> them on their side ? I've pointed them multiple times to our sql 
> scripts but it seams to fall on deaf ears.

Disclaimer: I'm not a member of the EPSG team, but I'm regularly in 
touch with an EPSG maintainer in the context of ongoing ISO 19111 revision.

EPSG is in a position between developers and mapping agencies. I do not 
know for 12.059a, but some issues reported as inconsistencies in 
previous versions were actually explicit demands from mapping agencies 
rather than errors. For example, there is the case of a missing epoch 
for the CRS of a northern country (I forgot the details, sorry). It was 
not an EPSG error, but a demand of that mapping agency who felt that in 
their particular case, providing a single epoch would be misleading and 
preferred to provide none rather than a misleading information. This 
case has been discussed at length during the meetings of the ISO 19111 
revision group a few months ago. The EPSG representative reported that 
they were struggling about how to conciliate both developer's grief and 
mapping agency concern in same time, and the group was considering 
whether a change of ISO 19111 model would help.

The vast majority of the debate in the ISO 19111 group is about how to 
improve our representation of the physical world, and whether a 
perceived "improvement" of ISO 19111 is really correct. After an 
agreement has been reached on geodesic grounds, the translation into 
data structures (UML of classes and attributes) tends to be quite 
consensual. While I do not know the internal of EPSG team work, I think 
that they also focus mostly on the geodesic content, maybe at the cost 
of compromises when a particular case does not fit nicely in the 
database schema. I think that the bar for changing the schema is also 
quite high, a little bit like changing a public API.

     Martin