[csw-maintainers] commentary on shared library naming proposal

[Dagobert Michelsen]

Hi Phil,

I just got an interesting example so I do post this myself now :-)

On Friday, November 19, 2010, Philip Brown <phil at bolthole.com> wrote:
> On 11/19/10, Dagobert Michelsen <dam at opencsw.org> wrote:
>> Hi Phil,
>> 
>> Am 19.11.2010 um 20:43 schrieb Philip Brown:
>>> Note, this is off-list, because I didnt want to needlessly extend an
>>> already huge thread, further, if not neccessary :)
>>> 
>>> 
>>> On 11/19/10, Dagobert Michelsen <dam at opencsw.org> wrote:
>>>> .....
>>> 
>> 
>> That means:
>> - CSWlibxyz may depend on CSWlibxyz-legacy containing legacy libs
>> as needed by existing packages
>> - CSWlibxyz may depend on CSWlibxyzM and CSWlibxyzN etc.
>> - CSWlibxyz must depend on the latest CSWlibxyzZ as it matches -devel
>> - CSWlibxyz must depend on CSWlibxyz-devel
>> 
>> In short: CSWlibxyz gets you everything. The name will also be used to
>> track bugs for every CSWlibxyz* package and libxyz/ should be the GAR
>> directory name.
>> 
> 
> sounds like we mostly agree. The only difference left is that I dont
> like the idea of literally HUNDREDS of empty "CSWlibxyz" packages. I'd
> like for them to be actually have some useful content.
> 
> seems to me like having them take on the role of "devel" for "current
> version of lib" is a useful thing for them to be doing.

I just finished libemf which has an interesting issue which I suggest
to split like this:
  CSWlibemf-devel contains the header files and *.so
  CSWlibemf1      contains libemf.so.1.0.0
  CSWlibemf       contains /opt/csw/bin/printemf as part of the lib package
Also CSWlibemf depends on both CSWlibemf-devel and CSWlibemf1.

Would this be satisfactory? Or better leave the (unnecessary) dependency
to CSWlibemf-devel and just depend on the latest runtime lib?


Best regards

  -- Dago