RE: Proposed Document Changes

["Hollenbeck, Scott" <shollenbeck@verisign.com>]

Roger,

I think the answer to your question could be found further down in the
message you quoted the text from.  Here's what it says about changes to the
TCP draft:

"- Eliminate/reword all of the text describing "asynchronous" operation.
Instead, it's OK to mention here that the client might be able to gain some
performance efficiencies by pipelining commands, but that pipelining doesn't
change the protocol's basic ping-pong operating mode."

The core protocol is and remains a command-response protocol, though TCP
(and maybe some other transports) have built-in features that would allow a
client to send more than one command before the first has been completed.
SMTP, for example, can work the same way: the client can buffer and send
multiple commands, anticipating that each will be performed correctly, but
the basic protocol is still command-response.

The <status> command can't return any information for a command that hasn't
been completed.  Nothing has changed about that.

-Scott- 

> -----Original Message-----
> From: Roger Castillo Cortazar [mailto:castillo@nic.mx]
> Sent: Monday, July 22, 2002 10:00 PM
> To: Hollenbeck, Scott; ietf-provreg@cafax.se
> Subject: Re: Proposed Document Changes
> 
> 
> 
> Hi everybody, I have a few questions about this changes 
> proposed by Scott 
> that could be interesting
> 
> >- All text describing "asynchronous" operation needs to be 
> removed/reworded.
> >Instead, text describing the protocol's operating mode as 
> "client-initiated
> >command, server response" needs to be reinforced.  Without 
> this change we'd
> >need to define our own "congestion control/ack" service to 
> handle the loss
> >of instances that may be queued in the application 
> queue-space (as when a
> >server goes down).
> 
> So, the "synchronous" operation mode is going to be mandatory ?
> 
> How is the command-response operating mode going to be reinforced ?
> - Is it that the client cannot send a command until he gets 
> the response to 
> the previous one ?
> - In other words: What happens if a client issues a command, 
> before he gets 
> a response to a previous one ?
> - Which response could he get if he issues a <status> for a 
> command, and 
> that command is not yet completed?
> 
> I think this could be somehow linked  to the <status> command and the 
> server's caching responsabilities.
> 
> Any comments ?
> 
> Greetings.
> 
> 
> Roger Castillo Cortazar
> NIC-Mexico http://www.nic.mx
> Top Level Domain .MX
>