Proposal to (formally/easily) allowing multiple versions of the same library installable
casey.jao at gmail.com
Fri Feb 20 12:47:10 UTC 2015
On 02/16/2015 08:17 AM, Ralf Corsepius wrote:
> On 02/16/2015 05:10 PM, Martyn Foster wrote:
>> On 16 February 2015 at 15:12, Kevin Kofler <kevin.kofler at chello.at
>> <mailto:kevin.kofler at chello.at>> wrote:
>> Christopher Meng wrote:
>> > Maintaining several version of the same library is not easy as
>> you think,
>> > basically once a developer wants to install version X while then
>> > people want to deploy things based on version Y, how to crack
>> this nut?
>> > You can't just care about runtime.
>> Then you need to patch one or the other package to work with the same
>> version. Only if that is not possible, a compatibility library can be
>> considered. But we should always first try to make everything work
>> with the
>> same version (if possible, the newer one).
>> The requirement to work with multiple versions of a package come up in
>> the scientific/HPC community very frequently. Its not always about API
>> compatibility, sometimes exact numerical reproduction is required which
>> isn't preserved even between minor versions (i.e. an OS update).
> I don't buy this argument wrt. Fedora.
> Fedora is a rapid moving, forward looking distro, in which such
> regressions should be fixed and not be worked around by compat-libs.
Since Fedora serves as a blueprint for RHEL, CentOS, and Scientific
Linux, which do get used in the scientific community and encounter the
issues Martyn mentioned, any technical framework that Fedora develops to
handle those issues would do the larger community quite a service, even
if it does not get used that often by Fedora users.
More information about the devel