[Haskell-cafe] Package Takeover: `toml`

[amindfv at mailbox.org]

On Fri, Mar 12, 2021 at 11:50:06AM +0000, Tom Ellis wrote:
> On Fri, Mar 12, 2021 at 12:36:16PM +0100, Henning Thielemann wrote:
> > On Thu, 11 Mar 2021, amindfv--- via Haskell-Cafe wrote:
> > > Again, trying to be respectful here, but "burning" kinda does imply
> > > "fire," and "need" certainly does imply "need." It's now seeming more
> > > just like a desire for the package name.
> > 
> > I have nothing to do with 'toml' but the many takeover requests in the
> > recent past make me nervous that if I am away from Haskell programming for
> > some weeks or months brings me in danger of losing my packages. Btw. for
> > some years I was not subscribed to Haskell Cafe because of high traffic and
> > I would have missed such takeover request. I think the preference should be
> > to create a fork.
> 
> This raises an interesting question: to whom does the entry in the
> package namespace belong? There's a tacit assumption that it belongs
> to the first person who registered it. Arguably though it could be
> deemed to belong to the community.  The more "generic" the name the
> more water that argument seems to hold.  
> 

As I said earlier, I'm open to this line of reasoning if there's a clear winner in terms of mindshare and functionality. As it is there are multiple popular, relatively mature packages, and the request was to give the name to a project which doesn't yet exist except in stub form.

"Belonging to the community" is a tricky concept, too - imagine the community speaks and declares tomland the winner, and tomland gets moved to the 'toml' namespace. Then, a couple years down the line, people find they want streaming parsing and in the meantime Carter and Emily have written great code. Do we then kick The Package Formerly Known As Tomland out of the 'toml' spot and put an entirely different project there? That's quite a breaking change for people with 'toml' in their .cabal files.

For what it's worth, my rewrite isn't a whole-cloth reimagning but instead is a fairly straightforward modernization of the existing 'toml' code.

Tom