John Hurliman wrote:
> It would be interesting to see how client-side encryption like the cache 
> would be successful with the open source client in 2007 that Cory 
> mentioned yesterday at the SL town hall.

It is possible to have both an open source client and a
"Trusted" main grid.

LL could set up two completely separate persistent grids
that would not share anything except for account information.

1. The "Trusted Main grid" would only send FullPerm
items to clients not compiled by LL.
  The LL "Trusted" client would have an "encryption class"
that would both encrypt the pipeline and the cache.

2. The "Open source" grid would allow any client to connect
and the "encryption class" would be an empty wrapper class.

A "proxy cache" could cache the items without being able
to actually know what they are if LL implements the
pipeline encryption at the object/item level.

> Databases were invented to solve all of these problems, maybe that would 
> be a better way to go?
> John

Requirements
  1. Muti-platform
  2. Free
  3. Fast
  4. Compiled in to application to prevent OS conflicts.
  5. Able to handle 1T byte on a low end PC (multiple drives)

Yes. A database is part of the task, but the question is to
use something that already exists or "roll your own" to
better tailor it to the task.

I got a bad taste with M$ Access.

No matter which database you use it will have the same problems.
The key to speed will be moving items to cluster them next to
each other based on usage in order to reduce hard drive seeking.

One suggestion on the forum was to use one directory per sim.
Even though that would store the same textures multiple times
it would be faster.

-- 
Bill, Grumble
_______________________________________________
Libsecondlife-dev mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/libsecondlife-dev

Reply via email to