[CF-metadata] Provide input on draft of STAC "Datacube" Extension

Ryan Abernathey ryan.abernathey at gmail.com
Mon Jan 21 04:16:00 MST 2019


Ryan,

That's a very good question. I'm not sure there IS a definitive advantage
of STAC over THREDDS. My main goal here is just to have some discussion,
since STAC is growing quickly in the adjacent world of geospatial imagery.
The conclusion may very well be, we don't need STAC.

Within Pangeo, our interest in STAC is in the context of putting netCDF (or
more accurately, netCDF-like Zarr) data into cloud storage as static
assets. We are looking for a way to catalog those assets, also in a static
file, and STAC is one thing we are considering. TBH, it never even occurred
to me that we could generate a static THREDDS catalog.xml file and use this
to describe our assets. In my mind, THREDDS catalogs are intimately bound
to opendap servers, and the THREDDS server in particular. But after reading
the THREDDS spec document you sent, I see that I am wrong about that.

Comparing the STAC vs. THREDDS specs, the main difference to me is the
relative high complexity of the THREDDS spec. I could pretty easy figure
out how to write a STAC catalog for my data in a text editor. I can't say
the same for THREDDS. But that is not a dealbreaker.

Beyond that, the STAC principles
<https://github.com/radiantearth/stac-spec/blob/master/principles.md> are
pretty appealing:

- Creation and evolution of specs in Github, using Open Source principles
- JSON + REST + HTTP at the core
- Small Reusable Pieces Loosely Coupled
- Specify in OpenAPI 3.0 (formerly known as Swagger) specification
- Focus on the developer. Specifications should aim for implementability
- Working code required. Proposed changes should be accompanied by working
code (ideally with a link to an online service running the code)
- Design for scale

The first one--evolution of specs in Github--is particularly important is
we explore new cloud-native approaches to data sharing. Perhaps the THREDDS
spec is indeed managed in such a way, but I couldn't find the repo. For the
second (JSON vs. XML), the THREDDS stack to me feels pretty java-centric in
general, based on an older generation of web technology. I don't want to
have to learn java. But again, I recognize it is possible to just use the
spec and not necessarily the software around it.

Best,
Ryan







On Sun, Jan 20, 2019 at 11:40 PM Ryan May <rmay at ucar.edu> wrote:

> Ryan,
>
> Can you enumerate some of the advances that STAC brings over the existing
> THREDDS catalog spec?
> https://www.unidata.ucar.edu/software/thredds/v4.6/tds/catalog/InvCatalogSpec.html
>
> From what I can tell, a lot of the concepts are similar, and the existing
> spec is already supported by tools beyond THREDDS, like Hyrax and ERDDAP. I
> think JSON is great and all, but I'm curious what adding a new standard
> brings us rather than extending an existing, already-supported standard?
>
> Ryan
>
> On Fri, Jan 18, 2019 at 7:23 AM Ryan Abernathey <ryan.abernathey at gmail.com>
> wrote:
>
>> Dear CF Conventions People,
>>
>> Some of you may be aware of the Spatio-Temporal Asset Catalog (STAC)
>> project.
>> https://github.com/radiantearth/stac-spec/
>> STAC is basically a .json specification which aims to standardize the way
>> geospatial assets are exposed online and queried. It originated from the
>> geospatial imaging community, and is described in this blog post by Chris
>> Holmes:
>>
>> https://medium.com/radiant-earth-insights/announcing-the-spatiotemporal-asset-catalog-stac-specification-1db58820b9cf
>>
>> There is currently some discussion on the STAC repo about how STAC could
>> be useful for netCDF-type data.
>> https://github.com/radiantearth/stac-spec/issues/366
>> This is a technology space that is currently occupied by THREDDS and
>> OpenDAP. But something like STAC could be very useful for our community,
>> especially as more netCDF-style data moves into the cloud.
>>
>> In particular, there is a proposed extension to STAC to describe "data
>> cubes," which is roughly what geospatial imaging people call netCDF-type
>> gridded datasets:
>> https://github.com/radiantearth/stac-spec/pull/361
>>
>> I wanted to email this mailing list to see if someone from the CF
>> community could weigh in on this proposed standard. More generally, if the
>> CF / netCDF community wanted to engage with the STAC people more broadly, I
>> think there is quite a bit of potential.
>>
>> Cheers,
>> Ryan Abernathey
>>
>> _______________________________________________
>> CF-metadata mailing list
>> CF-metadata at cgd.ucar.edu
>> http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
>>
>
>
> --
> Ryan May, Ph.D.
> Software Engineer
> UCAR/Unidata
> Boulder, CO
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cgd.ucar.edu/pipermail/cf-metadata/attachments/20190121/fb12f030/attachment.html>


More information about the CF-metadata mailing list