RE: Size matters was Re: notification charter proposal

["Hector Trevino \(htrevino\)" <htrevino@cisco.com>]

I agree that this is part of the protocol definition but I see it as
outside the scope of the charter definition. I'd assume none of this
will be included in the charter.


I don't think the spec should set a hard limit on the max message size.
I agree w/Juergen about not limiting the protocol but rather allowing
implementations to advertise their max message sizes (if one is set). It
may be better to say something like app layer message segmentation is
not allowed. But recommendations/guidelines for max message sizes should
be provided in the spec. 

Hector

    
 

-----Original Message-----
From: owner-netconf@ops.ietf.org [mailto:owner-netconf@ops.ietf.org] On
Behalf Of Andy Bierman
Sent: Tuesday, November 22, 2005 4:01 PM
To: Tom Petch
Cc: netconf@ops.ietf.org
Subject: Re: Size matters was Re: notification charter proposal

Tom Petch wrote:

><inline>
>Tom Petch
>----- Original Message -----
>From: "Andy Bierman" <ietf@andybierman.com>
>To: <j.schoenwaelder@iu-bremen.de>
>Cc: "Tom Petch" <nwnetworks@dial.pipex.com>; <netconf@ops.ietf.org>
>Sent: Tuesday, November 22, 2005 4:41 PM
>Subject: Re: notification charter proposal <snip>> >
>  
>
>>That would be fine with me.
>>I brought up the issue because the problem is obviously worse for 
>>notifications than RPCs, because the lack of an application layer ACK 
>>(which <rpc-reply> effectively is).
>>
>>The reason for setting a minimum is to allow NMS developers to code to

>>some reasonable minimum.  The market will help keep things somewhat 
>>reasonable, but we could see a range from 1500 to 64K as the minimum.

>>Is that OK?
>>
>>Andy
>>
>>    
>>
>
>I raised size because there are references appearing on the list 
>calling for something syslog-like and size has been an issue with 
>syslog that has not been satisfactorily resolved.  syslog unearthed two

>contrasting viewpoints, one that messages should not exceed 2k or even 
>1k, the other that 64k should be supported (for at least one
application).
>
>Netconf is more virgin territory than syslog is but I do believe is 
>setting a Maximum size, at least to start with; but I do not see this 
>as anything to do with the charter:-).
>  
>

This is part of protocol syntax and semantics.

IMO, the best approach is to have the NC peer advertise its
max-message-size.  The number should be an PositiveInteger, so that
would set a theoretical max-message-size of 4G-1 bytes.


A separate issue is whether there is a minimum an NC peer can set this
parameter to, somewhat higher than zero.  I don't see any consensus yet
on how much higher to set the minimum.


Another  important issue:  Can an NC peer (optionally?) send its
max-message-size in the <hello> exchange?  This doesn't mean the other
NC peer has to do anything with the value, it just provides a good guess
at the number instead of trying
various message sizes to see what breaks.   Advertising a
max-message-size doesn't mean you won't get messages greater than that
size. 


>Tom Petch
>
>
>  
>

Andy

>
>  
>


--
to unsubscribe send a message to netconf-request@ops.ietf.org with the
word 'unsubscribe' in a single line as the message text body.
archive: <http://ops.ietf.org/lists/netconf/>

--
to unsubscribe send a message to netconf-request@ops.ietf.org with
the word 'unsubscribe' in a single line as the message text body.
archive: <http://ops.ietf.org/lists/netconf/>