Abdulaziz Ghuloum <[email protected]> writes:

> On Aug 24, 2009, at 7:31 PM, Andreas Rottmann wrote:
>
>> The mapping can probably be the identity in many cases, however. If
>> there is no "natural" mapping, we could use <year>.<month>.<n>, or
>> <year>.<month>.<day>. A problem might be if there *is* a natural
>> (e.g. identity) mapping, but there are packages available from $VCS
>> as a
>> well. This complicates things -- you have to be careful so that the
>> version order relationship is kept. I can not see how to solve this
>> issue without resorting to a (slightly?)) more complicated versioning
>> scheme.
>
> A slightly more complex scheme can be dash-separated list of versions
> each of which is a dot-separated list of numbers (e.g., 12.3.4-0.1).
> So, you can use <year>.<month>.<day>-<rev> or even
> <relyear>.<relmonth>.<relday>-<revyear>.<revmonth>.<revday>.
>
> [...]

Fine proposal -- this buys us the most of the benefits of Debian's
(relatively complex) versioning without all the hair.


>> In the general case, yes. Upstream might have an insane numbering
>> scheme, but I think this is an exception, and most upstream release
>> numbers can be used with little or no mangling, if you relax the rules
>> for version numbers a bit (as Debian does).
>
> I think they would still be used with little or no mangling anyways
> (since the upstream packages are, for the most part, noninexistent
> yet :-))
>
Yes, there's just a few of them ATM, but hopefully that will change :-).

Regards, Rotty
-- 
Andreas Rottmann -- <http://rotty.yi.org/>

Reply via email to