[Haskell-cafe] Subcategories on Hackage

[wren ng thornton]

On 6/4/11 9:41 AM, Tillmann Vogt wrote:
> Am 04.06.2011 15:00, schrieb Sebastian Fischer:
>> http://www.shirky.com/writings/ontology_overrated.html
>
> I think I have read that article a long time ago. I just looked at it
> again. Very lengthy but true.
> In the middle there is a section when ontologies work well and ... it
> all applies to Hackage.

I'm not so sure of that.

> When Does Ontological Classification Work Well?
>
> Domain to be Organized
>
> Small corpus    3k libraries is nothing compared to the whole web
> (referring to the yahoo example), or every book on earth

Sure, it's small by library standards; but then again, libraries have 
curators. Libraries don't have to deal with everyone just filling out 
their own metadata when donating a book to the library. In comparison to 
others corpora of its kind, Hackage is of middling size and still 
growing. The relevant corpora here are things like CPAN, CTAN, CRAN and 
the like; not the whole internet, nor even a university library.

> Formal categories    already exist, just need some improvement

The categories aren't formal. Formal categorization means things like 
the names of biological species or the various constructs of abstract 
algebra, things which have a fixed and definite definition. Even the 
ontology of biology suffered a major upset over the last two decades as 
genetic sequencing demonstrated that the previous organization of 
species was woefully wrong.

When is something a 2D graphics library vs a 3D library? What about all 
the shades in between, all the various options like depth parallax, 
pseudo-3D, fixed perspective, sprites on 3D backgrounds,... before you 
enter true 3D rendering? Something like the raster category may be a bit 
more formal, but even with my feeble knowledge of graphics I know that 
graphics as a whole is a mish mash of differing things and not a 
formally specified domain.

> Stable entities    Haskell is a mathematical language, and if there is
> one thing that will never change then it is math

Haskell is quite mathematical, but mathematics have shifted and changed 
dramatically over the years. Math changes all the time.

Programming libraries and projects even moreso. Goals and directions 
change all the time. What began as a project in category A could easily 
shift to become focused on category B.

> Restricted entities    access only after personal email

What Clay means here is things like a library that only deals in bound 
volume books, as opposed to dealing also in other literary media 
(papyrus, velum,...) or in non-literary media (music, video,...).

This does describe Hackage pretty well though. It's only for software, 
and only for software written in (or for) Haskell. While that's still a 
very broad domain, at least all the entities in it are of the same kind.

> Clear edges ?
>
> Participants
> Expert catalogers    Maybe the best educated programmers in the world

Even if we are the best educated, I doubt there are very many Haskellers 
with expert-level experience in cataloging. How many of us work as 
librarians, ontologists, or other professions in charge of organizing 
and curating data? There are a few, I'm sure. Some work in databases, 
and some do annotation work (e.g., the NLPers), but the average 
Haskeller is not an expert cataloger. And I'm sure the experts are too 
busy to have all of Hackage foisted upon them.

> Authoritative source of judgment    Maybe

Hackage has no Benevolent Dictator For Life (that I'm aware of).

-- 
Live well,
~wren