Splitting Network.URI from the network package

[Michael Snoyman]

On Wed, Aug 13, 2014 at 2:35 PM, Johan Tibell <johan.tibell at gmail.com>
wrote:

> On Wed, Aug 13, 2014 at 11:39 AM, Michael Snoyman <michael at snoyman.com>
> wrote:
>
>> On Wed, Aug 13, 2014 at 12:28 PM, Johan Tibell <johan.tibell at gmail.com>
>> wrote:
>>
>  *Options 2: Users of Network.URI depend on both network and a specially
>>> crafted network-uri*
>>>
>>
>>>
>>>> ...
>>>
>>>
>>  Cons:
>>>
>>>    - network-uri has a false dependency on network (i.e. it doesn't
>>>    actually need that package).
>>>    - You can't build against a new version of text *and* use the
>>>       network-uri package (this is the current problem we have with network in
>>>       the problem description).
>>>
>>>  Can you clarify this point? I would imagine that network will no
>> longer have *any* dependency on text, so I don't see where this comes from.
>>
>
> My apologies. Let me clarify. If the user doesn't already have a version
> of network installed (e.g. via the HP), then building network is required
> to build network-uri. This probably isn't a problem on Windows, if we
> assume users already have an appropriate version of network through the HP.
> It might be an inconvenience (e.g. longer build times) for users who don't
> use the HP but still want to build network-uri (e.g for the maintainer of
> network-uri :) ).
>
>
Thanks for the clarifications Duncan and Johan. Yes, we should add a con to
the option 2 that usage of network-uri will require network to be
available. I'd consider this a relatively low-impact con, since I highly
doubt there are many people out there who will want to use Network.URI but
not also want to use network- at least transitively.

Even after the arguments from Duncan and Johan, I still would prefer going
with option 2, because (1) I don't feel confident yet that all flag-related
issues in the dependency solver have been fixed (up until just two weeks
ago I was still answering user questions about those bugs), and (2) my
experience with the flag approach was that it was very tedious to work
with, and I remember seeing a lot of confusion among other packages as to
the right way to specify dependencies.

I still think that either option 1 or option 2 are better than the current
status quo, so I'd rather not let this issue become a sticking point in the
proposal. I'd say let's take a vote on option 1 or 2, and continue with the
discussion deadline for this proposal (which seems to have unanimous
support) of this Friday. Any objection?

Michael
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.haskell.org/pipermail/libraries/attachments/20140813/9ca04553/attachment.html>