Initial comments

[Simon Reinhardt]

Hi,

First of all this service seems like a really good idea. For a project
where I need to let people find identifiers for things to link to I
was planning to search DBPedia, UMBEL and others and then merge the
results - but this would still produce duplicates. Open GUID will make
this a lot easier. Anyway, some comments.

The spec says oguid:identical is transitive, however the example on
http://openguid.net/specification#transitivity actually shows that
it's symmetric.

Then on http://openguid.net/specification#what it says that "As long
as something is determined unique by the OpenGUID.net community, it
can have an Open GUID." But what happens in cases where you detect a
"duplicate" afterwards and want to clean it up, because of this
constraint of uniqueness? According to the same section you can't
because "Once established, an Open GUID will always refer to the same
concept." So what's the community process of adding new GUIDs and how
does it ensure that no duplicate or otherwise odd things get added?
Seems like this would be very fragile and has to be treated with the
utmost care. That's something communities don't do. :-)

Regards,
Simon

[ja...@openguid.net]

Welcome Simon and thanks for the positive feedback.

Good eye! There were meant to be both symmetric and transitive
behaviors stated. I will fix it.

Good question on dealing with duplicity. I wrote a bit more about the
plans in the roadmap, but should add it to the spec.

Duplicates will merged in a way that allows both to mean the same
thing. Visually it will be a simple 301 redirect on the website to an
arbitrary primary GUID, with tags & relations merged. Semantically a
permanent rdf identical statement will be in the published linked
data.

The exact process for merging has not been nailed down yet. I'm
imagining a wikipedia-style editorial system, but more democratic.
The concepts should be considered the same for the majority of humans,
possibly with voting on the site. (random idea off the top of my
head...how about a poll widget placed on any site that asks 'Are these
identical'?). It will take some experimenting to get it right, but it
is a solvable problem, imho.

Thanks much for the help.

On Sep 25, 4:17 pm, Simon Reinhardt <simon.reinha...@gmail.com> wrote:
> Hi,
>
> First of all this service seems like a really good idea. For a project
> where I need to let people find identifiers for things to link to I
> was planning to search DBPedia, UMBEL and others and then merge the
> results - but this would still produce duplicates. Open GUID will make
> this a lot easier. Anyway, some comments.
>

> The spec says oguid:identical is transitive, however the example onhttp://openguid.net/specification#transitivityactually shows that
> it's symmetric.
>
> Then onhttp://openguid.net/specification#whatit says that "As long

[Simon Reinhardt]

On Sep 26, 12:04 am, ja...@openguid.net wrote:
> The exact process for merging has not been nailed down yet.  I'm
> imagining a wikipedia-style editorial system, but more democratic.
> The concepts should be considered the same for the majority of humans,
> possibly with voting on the site.  (random idea off the top of my
> head...how about a poll widget placed on any site that asks 'Are these
> identical'?).  It will take some experimenting to get it right, but it
> is a solvable problem, imho.

MusicBrainz has a really elaborate moderation system for that. Edits
need to be voted on or not, depending on the editor's status, the edit
type and the data quality of the target of the edit. Some other
aspects in the lifetime of an edit also get influenced by that.
Generally people are lazy to vote on edits but MusicBrainz uses
various tricks to increase the vote-rate.
So maybe you can use that system as a source of inspiration.
http://wiki.musicbrainz.org/VotingFAQ and http://wiki.musicbrainz.org/HowVotingWorks
have more information.

Simon