Hi Werner:

In message <20210322193839.Horde.X35YQqE4kZiE_wnnKiMRRjO@www.rxd.de>,
Werner Hunger writes:
>I think your changes are correct as such, but i'm not sure about
>back_mysql in total, and i did not do any testing yet. db_create() for
>example still contains "CHARACTER SET utf8" for "CREATE DATABASE", and
>i'm not sure whether a proper collation definition should also be part
>of database creation/configuration.

Understood. Would you recommend using the character set from the
config (mysql_charset) in the CREATE DATABASE?

>The problem is, our version of Roundup is a heavily patched 1.4.16 for
>which the MySQL-dependent code differs substantially from the upstream
>code, so i can't compare their behaviour without some effort. Our
>version of the backend has been changed to work around the n+1 problem
>for tree-like and other strongly linked data structures that we have
>in our tracker, and that is now up with data from 2004 on (about 1.5
>Mio messages). The original code (in combination with our page
>templates) generates way too much database queries for one web request
>- MySQL is not able to handle this load with decent performance.
>Therefor we added support for bulk node prefetching with node id lists
>anticipated from preceeding query results, among other things.

Hmm that sounds interesting. Might some of the ideas in:
https://issues.roundup-tracker.org/issue2550514 help reduce the need
for your changes?

>The reason i was looking into the current docs was to start estimating
>whether it would be feasible to migrate and merge our backend changes
>to Roundup 2.x - and thus gain Python 3 compatibility. If you'd like
>to, i can keep you posted once we decide to upgrade and the merge has
>been done.

I would very much like that. Also please consider joining and keeping
roundup-devel at lists.sourceforge.net in the loop.