Re: [Nel] A small document for your consumption

[Thierry Mallard]

On Fri, Apr 13, 2001 at 11:34:53AM +0200, Vincent Archer wrote:
> Steps
> -----
> [...]
> 5: The LS sends a notification to the selected WS of the client's connection
>    desires. It generates and submits a single-use cookie to validate the
>    incoming connection.

The LS <--> WS connection should be studied, perhaps ?
(if it wasn't intended in this document, then let's see that later.. ;-) )

[ *err.. ok i just read the end on the original mail, just forget it* ]

> 6: The WS selects a FES to accept the client connexion, and submits the cookie
>    to the FES.
> 
> 7: The FES acknowledges its capacity to accept the client to the WS.

So the WS is (or can be?) a load-balancer to all the FES in a given world ?
-- the balancing being done at network level, not process level --

> 8: The WS acknowledges its capacity to accept the client to the LS, and
>    indicates the IP/port of the selected FES.
> 
> 9: The LS acknowledges the login request to the client, and indicates the
>    IP/port of the selected FES.
> 
> 10: The client disconnects from the LS.
> 
> 11: The client initiates a connection to the indicated FES.
> 
> 12: The client sends the submitted cookie to the FES.
> 
> 13: The FES validates and acknowledges the cookie.

IMHO, as said in the other mail, the client should then disconnect from the
LS ; not before. The downside I see in this case would be the extended time of
connection (steps 11 et 13), which will lead to more network load. I don't see
how important that would be.

> Side notes
> ----------
> 
> Whenever a world starts, the WS establishes a permanent link with the LS,
> using an encrypted link (it is assumed that the LS and WS are located on two
> physically and probably geographically distinct networks). A 'SHARD' message
> serves as authentification, and the WS then updates the LS with its state,
> name and IP address. The WS may have a list of valid IP/port address for WS
> to avoid the occasional pirate server registration.

ok, so i should have read the whole document before arguing ;-))

Hope this helps..

-- 
Thierry Mallard              | http://vawis.net
GnuPG key on wwwkeys.pgp.net | http://erlang-fr.org (new)
key 0xA3D021CB               | http://worldforge.org