[csw-maintainers] A chain of updates

Peter FELECAN pfelecan at opencsw.org
Wed Dec 14 19:01:08 CET 2011


Dagobert Michelsen <dam at opencsw.org> writes:

> Hi Maciej,
>
> Am 14.12.2011 um 12:23 schrieb Maciej (Matchek) Bliziński:
>> 2011/12/14 Peter FELECAN <pfelecan at opencsw.org>:
>>> "Maciej (Matchek) Bliziński" <maciej at opencsw.org> writes:
>>>> Here's where named releases can help.  We could require sequential
>>>> updates. For example, legacy→dublin would be a valid update path, and
>>>> so would be dublin→kiel. However, legacy→kiel would not.
>>>> 
>>>> If we tracked the renames, we could identify all packages that e.g.
>>>> have gone 1→2 during the legacy→dublin transition, and have them go
>>>> 2→3 during the dublin→kiel transition.
>>>> 
>>>> Thoughts?
>>> 
>>> Question: does this means that I can do dublin -> ++kiel ?
>> 
>> You can do dublin→kiel, but you cannot upgrade from dublin to a named
>> release after kiel.
>
> ...and you cannot update from "legacy" (the old "stable") to "kiel",
> but must first update to "dublin".

This is clear now.

>> In other words, you need to sequentially upgrade
>> from one named release to another, with no skipping.
>> 
>> A thought: pkgutil could somehow keep track of this, and stop or warn
>> if someone attempted to do an invalid upgrade.
>
> +1.
>
>>> Also, this means that you need to keep all the named releases or define
>>> an obsolescence time for a named release.
>> 
>> Correct.  We'd keep them for something like 3 years, which amounts to
>> about 6 named releases at any given point in time.
>
> We could keep them on the primary mirror without expiration, but provide
> for the rsync downstream only that subset.

If you have enough place that would be the best, giving an long term
archive capability.

-- 
Peter


More information about the maintainers mailing list