[sldev] SL Database purge

[Felix Duesenburg]

Thomas Shikami wrote:
> Argent Stonecutter wrote:
>> ... . The operations on the SL database should prevent explicit loops
>> from happening, but they're also inferring references from the
>> contents of some assets, such as scripts, and database corruption can
>> create the damndest structures. So what do you do?
>>
>> Well, one common solution is to simply cut off the walk after a
>> certain distance. So...
> The other way would be to keep a list of assets already scanned and
> another list of assets that are in use. The first pass would be to fill
> the list of assets in use from inventory and sim database. Then there
> will be as many passes as needed doing the following, search of assets
> used that aren't scanned yet. Scan them through and fill the list of
> assets in use. Add each scanned asset to the list of assets already
> scanned. Repeat until the list of assets already scanned and the list of
> assets in use match. With this algorithm, even repeats won't do any harm
> and box in box in box in box is scanned to any arbitrary depth.


Wash, rinse, repeat... ;)

Ok, this is not meant to send LL recipes about how to do their job. That
would mean the same as sending the French recipes how to make a
croissant that doesn't crumble. But the guesswork comment about
serializing triggered an explanation, not meaning to call anyone naive
(sorry if it came across as such). It may have been misaddressed, but
with the intent of also getting to those who started this thread by
spreading badly researched rumours. Otherwise I would have replied off list.

I am getting the missing gesture thing, too. However, I /do/ think it's
unreasonable to suggest LL doesn't know how to do a proper search
through a system that they created themselves. There is obviously a bug
somewhere in the retrieval system, but it doesn't have to mean that all
is lost. Or maybe it is. Who knows.

What I find more disturbing is the fact that we ever had any case of
inventory loss at all that could not be resolved. If you have a
collection of assets that are worth anything to anyone, regardless
whether measured in terms of money or otherwise, then you /should/ have
a full audit trail recording all relevant transactions from the very
beginning. If it ever happens that a link between an object and the
information where/whom it belongs to gets lost, it can be restored upon
request. (Unless the audit trail itself gets lost... which leads us to
another kind of circular reference...) Such a system is obviously not in
place, or it's not functioning properly (yet). What about the
transaction history on the website, how does it look like at the
backend? Could it be used to also record rezzing/unrezzing or other
events, is that being considered or in development?

I'd really care to learn more about what are the actual facts here, and
I do believe that LL would get much less heat if this was addressed with
the same openness that is applied in other contexts, and which always
draws more praise than criticism even when things go wrong. Hope I
didn't miss anything I should have read, but if so then too many others
missed it as well.