[sldev] Re: Exciting reuter's post about OpenSim - project to create open source version of SL Servers

[Laurent Laborde]

On 9/8/07, Callum Lerwick <seg at haxxed.com> wrote:
> On Sat, 2007-09-08 at 06:11 +0200, Dale Glass wrote:
> > It also makes a lot of caching impossible as with no central server there's
> > no universal ID for an object (probably fixable by using content based
> > hashes)
>
> A content hash, of course.
>
> > Inventory becomes complicated, if there's no central server, where is it?
> > On the user's computer? And what if they log in from another box, or lose
> > their local copy?
>
> P2P-ish technology. Its nothing but a big distributed cache anyway. It
> will become as basic an internet service as DNS is. All ISPs will run a
> "seed" farm for its customers. Or at least contract it out to some
> really big seed farms. :)

As far as i know, Linden Lab is already using a distributed farm of
squid proxy cache. (squid was designed to be distributed).

They could be on sim server if they sim load wasn't so high.
You also have a local cache.

And P2P, like bittorent, do have some kind of central server too.
Bittorent is less centralized, and much more pain to find content.
It's not reliable like LL asset server, and our inventory certainly
need reliability.

What i want, really, is the abbility to setup a squid server at home
to cache all the content.
I certainly know how to setup a squid server at home, i have the
hardware, i just don't know how to setup everything to make it work
like a SL content cache.
And it's a must have for a company too... or even... an ISP ? (like
the good old internet with ISP running farms of proxy to save
bandwidth).

I'm sure it will happen. the question is "why isn't it happening already ?"

-- 
Ker2x