[Opensim-dev] User services refactoring status

[diva at metaverseink.com]

I'm neutral about keeping support for SQLite in core. For me, it's all a 
matter of someone taking the lead on maintaining it. Honestly, I agree 
with Justin, and that's the reason why I'm not doing that maintenance 
work myself. Having a no-installation OpenSim is very nice just to make 
it work and to have a feel for what OpenSim can do, but I think that the 
particular implementation of that concept -- with SQLite -- is quite 
deceiving. No one in his/her right mind would use SQLite as the DB for 
serious uses of OpenSim, just like no one uses it as the backend to web 
applications (there are always exceptions to every rule, of course!).

As such, starting off with SQLite, liking OpenSim, and then discovering 
that one needs to migrate the data from SQLite to another DB system is 
not situation that we should induce people into.

In order to implement the no-installation concept, which is a good 
concept, I'd much rather have an in-memory DB layer with no persistence 
at all. All data would disappear once the sim shuts down. That way it 
would be very clear to everyone that if they want persistence they need 
to think and research about which database system works for them. In 
some rare cases, people may come to the conclusion that SQLite is what 
works best for them. Those people should take the lead on maintaining 
the SQLite OpenSim connector.

Justin Clark-Casey wrote:
> Daniel Smith wrote:
>> As a developer of other things, but a user of OpenSim, I would say 'keep 
>> sqlite'.
>>
>> Keep the barriers to entry as low as possible.  If someone wants to try out
>> OpenSim, make it easy for them.  Put your end-user hat on for a moment.
> 
> I would usually make the same argument and this was why I did feel a preference keeping sqlite.
> 
> However, on reflection, I'm not sure whether it actually helps that much.  OpenSim is always going 
> to be somewhat complicated to set up and run so the added necessity to install a database as well 
> doesn't seem too onerous to me.  Mysql is a pretty simple install on all platforms.
> 
> We also quite often seem to get situations where someone has started off with sqlite and later 
> realizes that they need to convert everything to mysql in order to grow.  This can be an awkward 
> process.  In addition, we already have a situation where some things are not implemented in sqlite 
> (afaik, attachment persistence still isn't implemented there).  It might be better if we don't make 
> the end-user suffer these issues.
> 
> Anyway, as Diva said, all this really depends on people stepping up to make the necessary code 
> changes for sqlite.  If people do then it will remain in the codebase.  If nobody does then I think 
> that we may be better off removing it for now.
> 
>> cheers,
>>
>> Daniel
>>
>> On Mon, Jan 4, 2010 at 11:57 AM, Robert Martin <robertltux at gmail.com 
>> <mailto:robertltux at gmail.com>> wrote:
>>
>>     Just adding my thoughts given that XAMP is very easy to setup (heck it
>>     would be simple to just have a pre-setup XAMP module as an add-on
>>     download) sqlite is not really needed.
>>
>>     side question i have a diva install where the DB crashed midway during
>>     the first  migration stage do i need to dump the tables and start
>>     over??
>>
>>
>>     --
>>     Robert L Martin
>>     _______________________________________________
>>     Opensim-dev mailing list
>>     Opensim-dev at lists.berlios.de <mailto:Opensim-dev at lists.berlios.de>
>>     https://lists.berlios.de/mailman/listinfo/opensim-dev
>>
>>
>>
>>
>> -- 
>> Daniel Smith - Sonoma County, California
>> http://daniel.org/resume
>>
>>
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> Opensim-dev mailing list
>> Opensim-dev at lists.berlios.de
>> https://lists.berlios.de/mailman/listinfo/opensim-dev
> 
>