Pointer: JIRA'd feature: return of objects instantly/X minutes after owner absence

Greetings,

a new JIRA was created by me to which I wish to draw your attention to, be you a builder, land owner or plain traveller:

http://jira.secondlife.com/browse/SVC-2227

The issue describes a suggested additional option to the autoreturn feature of parcels and estates, with which objects not belonging to a group member or land owner would get returned instantly or after X minutes if the owner leaves the parcel or estate.

This would help with alot of problems, from cleanups to preventing spying or attacks, and could overall be used to reduce litter and improve performance.

Please take your time to read the JIRA, draw your friend's attention to it, and vote on the issue or leave comments and suggestions on the JIRA itself.

Thank you.

p.s. Apologies in advance if this is not the right board for this, but I couldn't find any better.

p.p.s. Links seem to parse in weird ways. Sorry for that.

Removed

Objects that are actively simulated (rezzed) are stored and maintained by the sim. When they're rezzed the sim does need to fetch the object from the asset server, but once it's rezzed the sim is responsible for it.

A return involves uploading the object back to the asset server and adding the new asset to your inventory on your inventory server and is rather "costly".

Your suggestion would at best keep the number of returns the same, at worst it would increase them and otherwise increase load on the asset servers (someone rezzes something on a sandbox, they crash, SL keeps them out for 5+ minutes, objects return when they otherwise wouldn't and when they log back in they rerez what was just returned which is an additional asset operation that wouldn't have happened otherwise).

(I do think it's a good suggestion, it's been discussed before, but saying that it would "improve" performance is a bit of a stretch when it would likely just add more load which doesn't make it very likely that LL will be interested)

Well, it was more in reference to client performance in littered sims, as well as script IPS in filled sandboxes :> Yes, returns would indeed upload back to the asset servers, but I doubt the amount would be really noticable on the system in general.

Kitty, the validity of your analysis depends on the times involved. I've worked in places where my objects were returned while I was working on them, so I had to rez them multiple times.

Assuming the *cost* of doing the checking is free, the OP's suggestion could increase or decrease the server load, depending on how it's used and how the land in question is used by folks to whom the auto-return policy applies. If set to at least 15 minutes on a sandbox, I think it would reduce both server and client load because the sandbox would generally get cleaned up faster after a builder leaves, reducing server load because there are fewer prims, and 15 minutes is usually enough time to relog after a crash, even if your whole computer crashes.

But the cost of this checking is probably not free. There might be a nice efficient way to implement it, and if so, I would be in favor of it.

If a specific sim is struggling, it only affects people on that sim, noone and nothing else on the grid will notice.
If the back-end services are struggling, all of SL is affected and everyone and everything starts to fail.

There likely is a (significant) cost to continually check all the objects on a parcel to see if they need to be cleaned up but that's again sim-local and doesn't affect anything but that single sim.

Suggesting that you should stress the infrastructure more to lessen the load on a sim is working backwards from a performance perspective so you can't use that as a "pro", it's a big "contra".

Why would there be a bigger load on the back-end services?

There is a bigger load on the back end if we have *more* object returns or takes. In a sandbox, every object rezzed gets either deleted or taken. The OP's suggestion increases the number of returns only in the case of a low timeout and crashes. It reduces the number of returns when the parcel's object return time is too short (when people get object returned while they're working on them).

Therefore, the back-end cost depends on the situation. Whether it would make a significant difference to back-end load is highly questionable. This is especially true if the lowest configurable value were 15 minutes, since that would avoid return/re-rez on crash, in the vast majority of cases. I think it would be a wash for the back end, so the server savings are significant.

I didn't, and wouldn't make such an ignorant suggestion. I very well understand scalability and performance considerations, with over 30 years of professional software engineering experience.