[OSGeo-Discuss] Tales from a Benevolent Dictator

[Rob Emanuele]

Apologies for veering off topic.

Hi Peter,

Thanks for citing your resources. Unfortunately I can't access the one
paper, since the only version I could find is behind a paywall, and the bar
chart you attached gives very little information; from these I cannot
understand the methodology or results. If you have more details I would be
happy to look further into this.

My concern is with your wide sweeping statements, and the implication that
rasdaman has been scientifically verified to be more performant than any
other system in all cases. This to me feels hyperbolically similar to
measuring that a bowling ball falls faster than a piece of paper when
dropped from the roof of a building and concluding that trees are the
objects which fall most slowly towards the earth.

For instance, I have doubts that those who had conducted the quoted
performance benchmarks set up the Apache Spark system in a way that
represents all potential configurations. I work on the GeoTrellis project
[1], which adds raster processing capabilities to Spark. I could for
instance imagine a system where raster data was stored in Accumulo, indexed
by GeoTrellis, and processed through Spark, which is very fast under many
query types. I won't make any assumptions on how fast as compared to other
systems, and it's very possible that rasdaman will beat out such a system
in a set of query types, or perhaps all queries. However, it is my opinion
that until the two systems were compared in such a way that everyone agreed
on on the methodology and the results, casually using the "fact" that one
system is "way faster" than the other system, and that one beats the other
"in all benchmarks" as an argument for some treatment from OSGeo (or for
any other purpose) deserves to be called into question, which I am doing
here.

I'd be happy to collaborate to develop, out in the open and in front of any
paywalls, an objective system of measuring performance between systems. At
which point in time we could make proclamations like, "[whichever
framework], under [these specific query types], running on [however many
nodes, whatever type of hardware], storing [this amount] of [this type of
data], performs better than [some other framework] under the same
conditions". Until then, I object to your very broad statements of
superiority.

Regards,
Rob

[1] https://github.com/geotrellis/geotrellis

On Sun, May 15, 2016 at 9:30 AM, Moritz Lennert <
mlennert at club.worldonline.be> wrote:

> On 15/05/16 14:40, Marco Afonso wrote:
>
>> Hi Anita,
>>
>> Aha! So there is a ponderation weight on software quality evaluation AND
>> project organization evaluation.
>>
>> So you can exclude an open source software with high quality if their
>> organization evaluation is low.
>>
>> For me that seems wrong. A software on a public repository is only
>> limited by it's licence terms, or unlimited at all. :)
>>
>
> But the discussion is not about whether the software should be in a public
> repository or not, or what the licence term should be. The discussion is
> about what the meaning of the "OSGeo project" label is.
>
> I don't think anyone has questioned the quality of the software, here.
> However, one of the aims of labeling a project an OSGeo project is to give
> a certain level of guarantee to potential users that this software
> _project_ respects a series of criteria that are considered important to
> ensure a long-term sustainability of that project. Putting one person's
> name in the statutes of a project and designating that person as the one
> who has ultimate decision rights (even if these decisions are always based
> on quality criteria), leaves the question of what would happen if that
> person lands under the proverbial bus.
>
> A more collective governance structure is seen by many as more sustainable
> in the long run. Similar debates have gone on for ages in Debian, for
> example, about team-based maintaining of packages vs individual maintainers.
>
> What I personally haven't really understood, yet, is what the rasdaman
> community is really afraid of. If the community works as well as described,
> why would the creation of a PSC-like structure create such problems ?
>
> Moritz
>
> _______________________________________________
> Discuss mailing list
> Discuss at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/discuss
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/discuss/attachments/20160515/82d7789f/attachment-0002.html>