On Thu, Sep 16, 2010 at 5:03 PM, James Tucker <jftucker / gmail.com> wrote:
>
> On 16 Sep 2010, at 11:57, elise huard wrote:
>
>> On Thu, Sep 16, 2010 at 4:45 PM, James Tucker <jftucker / gmail.com> wrote=
:
>>>
>>> On 16 Sep 2010, at 06:54, elise huard wrote:
>>>> - you could just reference a certain number of gems from ruby config
>>>> files, and the desired/tested version. this way the author can
>>>> continue work independently of ruby-core, and ruby core can also
>>>> approve or certify a certain version for a next release. =A0There's no
>>>> reason to impose extra versioning IMHO.
>>>
>>> This is not true. Users will do `gem update` and shoot themselves in th=
e foot. As one of the people who spends a lot of time on user support, I wo=
uld be really really upset to see this kind of dangerous action. As Marcus =
points out, there are even issues with this as an approach for versioning b=
etween different rubies too. We can protect against a /portion/ of this wit=
h the required_ruby_version flag in the gemspec, but I at present I /believ=
e/ that's only a minimum version marker. For this purpose it will need '=3D=
' or at least '~>' semantics or options to be sane.
>>
>> =3D means requiring an exact version of a gem.
>> http://docs.rubygems.org/read/chapter/16
>
> Exactly.
>
OK, well I think I misunderstood your response.  I meant, meaning an
exact version, and I don't really see the problem with that.