[sldev] Opening the server source?

Adam Frisby adam at gwala.net
Mon Jul 2 16:10:09 PDT 2007


I think the best focus there is to try keep everything as standard as 
possible, for these central services your not gaining much by putting 
another server in the middle other than read-only not-regularly-updated 
data, so keeping the interfaces along HTTP and using a configured proxy 
like squid would seem ideal.

In terms of scripting / physics engines - you still need to be able to 
upload and compile, and once you are allowing that, you open a big can 
of worms with security, doing it without opening that would be a very 
interesting challenge.

I do agree it might be a while before we see the official simulator open 
though. I'm betting 12-18+ months myself.

Adam

Mike Monkowski wrote:

> Adam,
> 
> I guess "message server" was a poor choice of terminology.  Perhaps 
> "transaction server" would be more descriptive.  The transaction server 
> would do the processing that the current sim server does to generate 
> messages to the other servers.  Some may be simple pass-through, but the 
> current "trusted" processing that the current sim server does could 
> still be done on LL transaction servers.
> 
> Having specialized per-service proxies leaves you with concurrency and 
> rollback problems, leaving more computation in the viewer, which of 
> course, is not trusted.
> 
> Yes, the scripts may be visible to the physics server, but scripts are 
> code, and, well, this is an open source project.  If you choose, you can 
> disallow your main grid scripts from being exported to private sims and
> vice versa.
> 
> Open sourcing just the physics and scripting could generate a lot of 
> creativity.  I'd hate to have to wait for a complete open source server 
> before being able to do anything.  That could be a very long time.
> 
> Soft Linden said,
> 
>> 2) If you can point to specific modules of the sim that you'd like to
>> see and can help me with an argument for getting them open sooner, I'm
>> all ears.
> 
> I'm asking for physics and scripting.
> 
> Mike
> 
> 
> Adam Frisby wrote:
> 
>> ...having a server manage connections for the client strikes me as 
>> somewhat unneeded complexity since all it is doing is proxying the 
>> connections; which could be served better by having specialised 
>> per-service proxies.
>>
>> Adam
>>
>> Mike Monkowski wrote:
>>
>>> Actually, I was suggesting something between these two.  By breaking 
>>> the sim into a message server and a physics server, you don't have to 
>>> change the rest of the architecture.  The viewer sees no change.  The 
>>> specialty servers see no change.  One box is just split into two with 
>>> the rest of the connections remaining intact.  It should be a lot 
>>> easier to implement.
>>>
>>> Mike
> 
> _______________________________________________
> Click here to unsubscribe or manage your list subscription:
> /index.html
> 



More information about the SLDev mailing list