Axel Donath

During the energy dispersion check for the event sampling we noticed the migra axis default we currently have is undersampling the energy resolution of CTA, but probably HESS as well,...

I think at some point we discussed introducing instrument specific default values in Gammapy for the IRFs. I still think this is a good idea. For now I would propose...

@bkhelifi, @AtreyeeS, @registerrier please comment!

@registerrier Yes, independently of this issue we can easily introduce such as mechanism, I'm just not sure whether we really need it. I think we already integrate spectral model using...

An alternative pattern could be to extract the info from the DL3 IRFs. ```python3 irf_spec = Observation.default_dataset_irf_spec dataset = MapDataset.create(geom, **irf_spec) geoms = Observation.default_dataset_geoms dataset = MapDataset.from_geoms(geoms=geoms) ```

While this is still an important topic, it's not a blocker for v0.19. We adapt the milestone.

Default true energy axis should be given with `node_type="center"`, however as long as the energy binning is fine enough it does not really matter (see #1967).

One proposal to infer an IRF axis binning from a set of IRF is here: https://github.com/gammapy/gammapy/pull/3711

Idea from dev meeting today: make behavior consistent with IRF kernel extraction. So find the nearest valid position to extract the exposure from. I think for a TS map we...

Thanks @QRemy, just a quick comment: in the past we also had https://reproject.readthedocs.io/en/stable/ as a dependency. If it provides the functionality (flux preserving reprojection) we can also easily re-introduce it...