[Opensim-dev] Release 0.7 RC1 Test Feedback

[Melanie]

Actually, that code is on my list. However, since there is such a
hue and cry over the new driver, I have refrained from putting into
master, since that would "freeze" the new driver in place.

Although I'm a proponent of forward motion, we MAY have to roll back
the driver and migration change. I really want to avoid that, but
until that issue is resolved, I just can't commit any code that
builds on it.

Melanie

Alex Ra wrote:
> DC> As far as I understand it, the new driver was added because it
> DC> supports stored procedures [better?]; stored procedures will allow
> DC> us to speed up some things. But 0.7 still doesn't have any stored
> DC> procedures, so the new capabilities are mute.
> 
> 
> Hello Diva,
> 
> It was me who put the updated MySql driver into the project, so I'm
> the one to kick for the mess.
> 
>  The reasons were:
> 
> (a) Yes, the ability to *define* multi-statement stored procs. MySQL has
> a rather hairy way to switch DELIMITER while defining an SP, so
> defining anything with a ';' inside was either impossible or very tricky
> with the previous driver.
> 
> You are also correct that stored procs are NOT used currently, however
> they can be used to provide "fast" tracking of asset access times,
> which, as I understand, is required for cleaning up unused assets and
> preventing the database from growing infinitely. The relevant code is
> ready, tested and described by me somewhere on Mantis, with a link to
> github, but I don't think anybody has bothered to review / merge it
> (yet).
> 
> The last time I looked, the access tracking functionality was
> commented out in the asset store code, because sending an UPDATE after
> each asset fetch just kills the performance (like 10 times slower).
> The solution based on stored procs has a performance penalty of about
> 10-15%.
> 
> I can re-post the link to Mantis where this approach is described, if
> anybody is interested.
> 
> (b) the less important reason is just to keep up with the most recent
> versions of the driver.
> 
> Updating the driver also involved some changes in the stores to
> correctly convert between values returned/accepted by the driver and
> UUIDs in the code. The reason is that the new driver implicitly
> converts CHAR(36), returning it as Guid (rather than a string). With
> the "Old Guids" config var, the driver would instead convert
> BINARY(16) to Guid.
> 
> With the new driver and the corresponding code changes in place, the
> "Old Guids" config var should NOT be used.
> 
> 
> Best regards,
> 
> Alex
> 
> _______________________________________________
> Opensim-dev mailing list
> Opensim-dev at lists.berlios.de
> https://lists.berlios.de/mailman/listinfo/opensim-dev
> 
>