[aur-general] TU application

[Connor Behan]

On 20/07/13 09:53 AM, Sébastien Luttringer wrote:
> On Sat, Jul 20, 2013 at 3:21 PM, Dicebot <public at dicebot.lv> wrote:
>> It is nothing of real importance but question is, what harm does
>> explicit versioning do?
> Interesting question because it underlines some benefits of being a
> rolling release for their maintainers.
>
> 1) Productivity
> You don't need to modify your package deps each time you update it.
> You only have to check (most of time, read the changelog).
> When upstream check this for you, in a configure or by  providing a
> test suite, you could even forget the minimal version check and free
> your mind to do interesting things.
> Arch maintainers manage a high number of packages, every unnecessary
> update is time lost.
>
> 2) Correctness
> More than avoid unnecessary changes in the update process, it avoid
> mistakes (and their fixes).
> When upstream doesn't provides minimal deps, it's long (and not easy)
> to find which minimal version is really required.
>
> 3)  Troublemaker
> The version is about a package, *NOT* an upstream source.
>
> I remember a discussion with a Debian developer about opening a bugfix
> because the required dep version is badly too high.
> A maintainer, wants a lib compiled with an option (e.g. --with-caca),
> so he bumps the lib version (e.g. from 1.15-1 to 1.15-2)  and require
> it in his package (e.g: from 1.10-1 to 1.15-2).
> The old versions (from 1.10 to 1.15 in the previous example) of the
> library is working perfectly (if you rebuild the package with the
> option --with-caca).
> Its happens (sometimes) when you do a backport on Debian.
>
> You don't know, by looking at it, if a required version is due to
> upstream or downstream change.
These are good reasons why a package maintainer should not obsess over
finding the minimum required versions for everything. But if you already
happen to know what they are, users who recompile packages or do other
customization would appreciate this information.

> 4) Speed
> It avoid pacman to checks version for each deps. This save a lot of
> useless computing (parsing and comparing two version)[1].
> Even if it's not a big deal on last intel processor, this is
> noticeable on slow processor (like raspberry pie, alix or soekris).
>
> We can resume these 4 points by saying : It's more simple.
>
> Of course there is drawback for people not updating the whole system.
> It's unsupported.
>
Those processors are no more supported than partial updates.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 555 bytes
Desc: OpenPGP digital signature
URL: <http://mailman.archlinux.org/pipermail/aur-general/attachments/20130720/04519a49/attachment.asc>