On 7/10/07, Thorsten Leemhuis <fedora(a)leemhuis.info> wrote:
On 10.07.2007 17:59, Patrice Dumas wrote:
> On Tue, Jul 10, 2007 at 08:17:46AM -0500, Rex Dieter wrote:
>> So, I was leaning toward following rpmforge's fftw/fftw3 naming scheme,
> Although naming fftw3 the newer version seems acceptable to me,
> especially if upstream breaks api now and then, naming the old version
> fftw seems wrong to me. [...]
>From looking at it at this point of time I tend to agree.
Normally I'd say that the latest version of "foo" should always be foo
as long as upstream calls it foo (IOW: fftw3 in this case is confusing
as upstream calls it fftw). If we ship foo (major-1) I'd expect we'd
call it compat-foo or foo(major-1) (e.f. fftw2 in this case). Calling
the older one simply foo (fftw) would IMHO be to confusing, as a lot of
people that simply want to install "foo" will expect the latest version
when they run "yum install foo".
Isnt there something about dropping the compat naming scheme and going
to something else? Or am I confused about a different email thread.
Here are my best attempts at Solomon
The first question I would ask is what are your customers wanting?
What packages need a Fast Fourier Transform and what do a majority of
customers running on EL want it to be? If the scientific community is
using lots of fftw2 versus fftw3 then that might be a good reason to
keep to older name schemes.
The second question is what is the new packaging name scheme? If the
name scheme is still compat-<package>-<major-minor> I would go with
that.. if it isnt then I would go with something like: fftw_22 fftw_3x
as the name schemes
Third I would go for an open and documented reasoning document from
both forge and you on why the names are different and how a user would
be able to deal with this issue.
--
Stephen J Smoogen. -- CSIRT/Linux System Administrator
How far that little candle throws his beams! So shines a good deed
in a naughty world. = Shakespeare. "The Merchant of Venice"