Re: cvs pserver performance

[Larry Jones]

Donald Sharp writes:
> 
> That's not necessarily true, there is not necessarily a need 
> to do it this way.  Sockets form tuples( client ip, client port,
> server ip/port )of information that provide information to the kernel 
> where to route the incoming and outgoing packets.  inetd in this
> case is managing the incoming data and sending the information to
> the correct cvs process.

You just contradicted yourself.  It is, in fact, the kernel that routes
network packets to the correct socket.  All inetd does is listen for
conections, accept them, and then hand them off to a server.  Accepting
a connection creates a brand new socket which inetd gives to the server
as it's stdin, stdout, and stderr and then closes (the server keeps it
open, of course).  After that, it's completely out of the loop; the
kernel delivers data to the socket and the server processes it without
any involvement by inetd at all.

> remember directory contention via lock files could also be significantly
> slowing down updates as well.  Especially over such a large repository.

If everyone is doing update, they're read locks which are sharable, so
there shouldn't be any significant contention.

-Larry Jones

I wonder what's on TV now. -- Calvin