[csw-maintainers] reminder about versioning, numbering, etc. The IPS factor.

[Peter FELECAN]

Dagobert Michelsen <dam at opencsw.org> writes:

> Hi Phil,
>
> Am 20.12.2010 um 21:06 schrieb Philip Brown:
>> FYI, a reminder about a "past" issue, that came to my attention again:
>> 
>> In the past, I only allowed "pure numeric" versioning for packages.
>> The Spec was
>> 
>> softwarename-##.##.##,REV=anythinghere.
>> 
>> a year or so ago, people pushed to drop the pure numeric constraint.
>> No-one else wanted to keep the pure numeric restrictions. So we now
>> have alphas to the left of REV, where only #.#.# used to be.
>> 
>> However, when and if we transition to integrating with "IPS", it
>> should be mentioned:
>> 
>> **they only allow pure numerics for versioning**
>> 
>> The archive thread at
>> http://comments.gmane.org/gmane.os.solaris.opensolaris.pkg.general/24163
>> somehow hasnt seem to have caught up with the latest, but here's a
>> quote from a later email from Danek Duvall, on the 17th Dec:
>> 
>> "I think it gets harder when you don't have a tuple of integers.  For
>> instance, OpenSSL uses letters in their normal versions -- 0.9.8j, for
>> instance, which we translated to 0.9.8.10. [....]"
>
> Umh, this is pretty bad. I suggest using instead
>   a.b.c_def,REV=x.y.z
> a version of z.y.x.a.b.c and hide def somewhere.

it wasn't in the REV string that we "hide" def? such as
a.b.c,REV=def,x.y.z,  in the case of openssl 0.9.8,REV=j
-- 
Peter