[Haskell-cafe] Package Takeover: `toml`

Sven Panne svenpanne at gmail.com
Fri Mar 12 15:31:32 UTC 2021

Am Fr., 12. März 2021 um 14:29 Uhr schrieb Imants Cekusins <imantc at gmail.com

> How about naming all official / recommended / _the_ packages with a
> prefix / suffix (e.g. base-*) and requiring an approval to create such
> packages?

I highly doubt that any finite (let alone: small) group of people has the
competence to decide what "the" package for a given task should be, given
the vast number of topics packages cover. It is OK for mainstream topics,
but even then different people have different needs and views. What happens
when people not really competent in a given topic try to standardize things
as "the" way to do it can e.g. be seen in C++'s SG13, a completely failed
attempt to standardize 2D graphics. Apart from a relatively small,
undisputed set of things, let the community decide what "the" way to do
things should be, basically using "survival of the fittest". If one library
is definitely better than another, then it will be used much more often, at
least most of the time. An e.g. well-curated "official" overview of
libraries for different topics can help here.

I think that discussions about package names are quite irrelevant, it is
all about discoverability of a package, and the package name doesn't help
there at all most of the time. Googling "haskell toml", you get tomland and
htoml as the first 2 hits. I would have never guessed the first name BTW,
and I actually don't care that much: Even if it's called "gnlpft" and it is
the 1st hit on Google and does its task well: So be it! Typing whatever
package name into a .cabal file is the least of your problems when choosing
a library. Another good example: "aeson". It's not really the first name
coming to your mind when you think about JSON, but people don't have a
problem discovering it.

A more problematic thing than the package names IMHO is the choice of names
for the hierarchical modules within a package: If things somehow clash by
accident here, you have bigger problems. There is no process whatsoever (at
least I don't know one) how these names are allocated. There were some
proposals by Malcolm W. and Simon M. some 10-20 years ago IIRC, but these
were only rough sketches.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/haskell-cafe/attachments/20210312/cd377a25/attachment-0001.html>

More information about the Haskell-Cafe mailing list