Re: Bug#732878: Add MariaDB as an alternative dependency
On 12/28/2013 11:10 PM, Clint Byrum wrote:
> Excerpts from Philipp Kern's message of 2013-12-28 06:23:31 -0800:
>> On Fri, Dec 27, 2013 at 06:36:44AM -0800, Clint Byrum wrote:
>>> So we have forks. And forks suck. But that is how MySQL's little inbred
>>> family works. And that is why I am pretty adamant that upstreams be
>>> involved or I will not spend what little time I do have for Debian on
>>> keeping their forks in Debian.
>> […]
>>> I'm not sure you can make a distributed database solution the same
>>> as a browser, which does not need to directly share and serve data in
>>> real time across nodes with other browsers as a primary function. The
>>> reasons for these forks are not mostly political like libreoffice
>>> vs. openoffice. There are deep technical differences that matter a lot
>>> to the users and developers of each fork.
>>
>> OTOH every introduction of a new fork will increase the burden on us to
>> support this "solution" for its technical differences. People will tell
>> us that we cannot remove fork X because their data is all in the format
>> of fork X. If they move away from being drop-in replacements, this
>> will become quite annoying I think (e.g. eglibc is a drop-in
>> replacement, egcs was basically a drop-in replacement, etc.).
>>
>
> Totally agree that once they go from drop-in replacements to full forks
> the situation gets more urgent for one to dominate and thus kill off
> the others. It seems to me that MariaDB 10 will force that situation,
> as it may introduce on-disk incompatibility. When it does, then we will
> have to get rid of some of the breaks/replaces relationships and move
> data storage from /var/lib/mysql to /var/lib/mariadb.
>
> To be clear, my point isn't to be anti-MariaDB or pro-MySQL. I am
> pro-Debian-user. Debian users are not served if we push them onto MariaDB
> when MySQL remains viable and its main code contributor (Oracle) remains
> committed to helping maintain it in Debian. Likewise, some MariaDB users
> have stepped up and thus it is also going to land in Debian. And Percona
> is working hard on getting their packages ready for Debian as well.
>
> So my point is, who are we to choose one if the authors and users of
> others are excited to maintain the others?
I totally agree that the situation sux for everyone: final users, people
willing to contribute to MySQL [1] server (they have to contribute to 4
projects now?), the Debian MySQL [1] maintainers, and finally anyone who
maintains something that uses MySQL [1] (package maintainers or upstream
who simply use the client).
All this is quite sad.
It'd be nice if there was ways to avoid some of the collateral damage
we'll have in Debian. Like avoiding to force adding mariadb-client in
all MySQL-using package would be a nice thing to do. Avoiding to first
do mysql-client | mariadb-client, and THEN migrate to the virtual
package would be even better. I'm just not sure what's the solution here
though... :(
Cheers,
Thomas
[1] Please understand MySQL at large in this context: I really mean
every forks of it.
P.S: Thanks Clint, for your valuable contributions to this thread.
Reply to: