
On Thu, Jan 19, 2012 at 12:34 PM, Felipe Almeida Lessa
On Thu, Jan 19, 2012 at 3:53 AM, Michael Snoyman
wrote: So the question is: how should future development continue? I see three choices:
1. Continue as-is. New modules can be added to these two packages, and I'll just route inquiries to the appropriate person. One minor change I'll insist on here is that authors specify in the module docs who the maintainer is, and I'll add them as contributors on Github. 2. No new modules will be accepted into these packages, they'll need to go into their own packages. 3. Even existing modules for which I'm not really the maintainer will be split off into their own packages.
For options (2) and (3), we can update the docs for authenticate and yesod-auth to point to external packages that provide additional auth backends.
I don't have any strong opinions, but I'd say (3). Note that yesod-auth could become a "meta"-package like the yesod package itself.
Cheers,
-- Felipe.
Just to clarify on choice (3). For the backends that I'm still actively maintaining (OpenID, BrowserID, etc), I don't really want to split them up into separate packages, since: 1. They all have the same basic dependencies (mostly http-conduit) 2. It's a pain for me to have to maintain multiple packages 3. I've gotten requests from the distribution maintainers (Debian in particular) to stop exploding the dependency list for yesod. So even under (3), I would want to keep most of the modules that are currently in authenticate and yesod-auth exactly where they are. I'm really talking about moving out the following modules: * Kerberos * OAuth * Possibly rpxnow: I don't use it actively anymore, but it hasn't really changed at all * Possibly Facebook, since there's a shiny new package[1] that might be a better home to its functionality ;) * Possibly HashDB (yesod-auth only) Michael [1] http://hackage.haskell.org/package/fb