Re: [netmod] syslog-model-11 single buffer vs list

2016-11-15 Thread Clyde Wildes (cwildes) 
Hi Jason,

I presented the latest draft ietf-syslog model last night in the Netmod session 
and mentioned that you had requested that I restore buffers to a list. No one 
raised any objections so I will do that in the next model update.

Thanks,

Clyde

From: "Sterne, Jason (Nokia - CA)" 
Date: Monday, November 14, 2016 at 7:58 PM
To: "Clyde Wildes (cwildes)" , "netmod@ietf.org" 

Subject: RE: syslog-model-11 single buffer vs list

Hi Clyde,

Most implementations probably have limits in the # of files, remote 
destinations, etc they will support.  If vendors decide to augment the model to 
add max-elements then they'd do it for a number of the lists here.  That 
doesn't seem like a big deal.  But trying to fit into a model with 1 buffer 
would require complete augmentation of the entire buffer destination.  So if we 
do keep buffer I really think it should be a list.

On the other hand I'm fine with removing both buffer and user sessions (similar 
reasoning for both -> they are supported by 2 vendors and each does it 
differently). Then vendors can just augment for those types of destinations.  
The other types have broader support/applicability.

Here is how we left off with the table (view this with fixed width font):

>Feature  Nokia   Brocade  Ciena  Cisco IOS/XE  Cisco 
> IOS/XR  Cisco NXOS  Juniper JunOS  Linux Rsyslog  Comments
>log-input-transports   
>   x
>log-action console xx x  x 
>  xx   x
>log-action buffer  x  x  x
>log-action filex  x  x 
>  xx   x
>log-action remote  xx   x x  x 
>  xx   x
>log-action terminal   x  x 
>  xx
>log-action session x   
>   x
>feature buffer-limit-bytes   x  x
>feature buffer-limit-messages  x
>feature file-limit-size   x  x 
>   x
>feature file-limit-durationx  x
>   x
>feature
> terminal-facility-device-logging
>feature
> session-facility-user-logging 
>x
>feature select-sev-compare x  x
>   x
>feature select-match   x  x
>   x   x
>feature structured-data
>   x   x Required because of RFC 5424
>feature signed-messages
>   x Required because of RFC 5848
>

Regards,
Jason

-Original Message-
From: Clyde Wildes (cwildes) [mailto:cwil...@cisco.com]
Sent: Tuesday, November 15, 2016 1:43
To: Sterne, Jason (Nokia - CA) ; netmod@ietf.org
Subject: Re: syslog-model-11 single buffer vs list

Hi Jason,

Buffer was a subject of discussion on the netmod list most recently by Tom 
Petch who raised some questions. In an e-mail on 2016/5/6 Tom said:

“The description of log-buffer confuses me.  The buffer is circular in nature 
so there is only one of them; but it is a list keyed on 'name' so there are 
lots of them.  This leaf configures the amount number of log messages that can 
be stored in the local memory logging buffer, so there is only one of them. 
Or?”

In the same e-mail Tom also commented on the complexity of the current model:

“My comment was more on the complexity that results from having so many 
options. Other models are worse - some of the routing ones I find 
unintelligible as a result - but I raised the issue on this model because it is 
being discussed on this list where (almost) all the expertise in these matters 
resides to see if anyone else would bite.”

In a reaction to Tom’s comments, I tried to simplify by changing the buffers 
list back to a leaf in draft 09. In retrospect this was a mistake. Note that 
AFAIK buffer is currently implemented by only two vendors: Cisco and 
Alcatel-Lucent-Nokia.

The Cisco implementation has one buffer and specifies the limit as the total 
buffer size in bytes.

The Alcatel-Lucent-Nokia implementation has multiple buffers and specifies the 
limit in total messages.

If we make buffers a list, we still have three features in the model and the 
necessity for implementations that support only one buffer to augment

Re: [netmod] syslog-model-11 single buffer vs list

2016-11-14 Thread Sterne, Jason (Nokia - CA) 
Hi Clyde,

Most implementations probably have limits in the # of files, remote 
destinations, etc they will support.  If vendors decide to augment the model to 
add max-elements then they'd do it for a number of the lists here.  That 
doesn't seem like a big deal.  But trying to fit into a model with 1 buffer 
would require complete augmentation of the entire buffer destination.  So if we 
do keep buffer I really think it should be a list.

On the other hand I'm fine with removing both buffer and user sessions (similar 
reasoning for both -> they are supported by 2 vendors and each does it 
differently). Then vendors can just augment for those types of destinations.  
The other types have broader support/applicability.

Here is how we left off with the table (view this with fixed width font):

>Feature  Nokia   Brocade  Ciena  Cisco IOS/XE  Cisco 
> IOS/XR  Cisco NXOS  Juniper JunOS  Linux Rsyslog  Comments
>log-input-transports   
>   x
>log-action console xx x  x 
>  xx   x
>log-action buffer  x  x  x
>log-action filex  x  x 
>  xx   x
>log-action remote  xx   x x  x 
>  xx   x
>log-action terminal   x  x 
>  xx
>log-action session x   
>   x
>feature buffer-limit-bytes   x  x
>feature buffer-limit-messages  x
>feature file-limit-size   x  x 
>   x
>feature file-limit-durationx  x
>   x
>feature
> terminal-facility-device-logging
>feature
> session-facility-user-logging 
>x
>feature select-sev-compare x  x
>   x
>feature select-match   x  x
>   x   x
>feature structured-data
>   x   x Required because of RFC 5424
>feature signed-messages
>   x Required because of RFC 5848
>

Regards,
Jason

-Original Message-
From: Clyde Wildes (cwildes) [mailto:cwil...@cisco.com]
Sent: Tuesday, November 15, 2016 1:43
To: Sterne, Jason (Nokia - CA) ; netmod@ietf.org
Subject: Re: syslog-model-11 single buffer vs list

Hi Jason,

Buffer was a subject of discussion on the netmod list most recently by Tom 
Petch who raised some questions. In an e-mail on 2016/5/6 Tom said:

“The description of log-buffer confuses me.  The buffer is circular in nature 
so there is only one of them; but it is a list keyed on 'name' so there are 
lots of them.  This leaf configures the amount number of log messages that can 
be stored in the local memory logging buffer, so there is only one of them. 
Or?”

In the same e-mail Tom also commented on the complexity of the current model:

“My comment was more on the complexity that results from having so many 
options. Other models are worse - some of the routing ones I find 
unintelligible as a result - but I raised the issue on this model because it is 
being discussed on this list where (almost) all the expertise in these matters 
resides to see if anyone else would bite.”

In a reaction to Tom’s comments, I tried to simplify by changing the buffers 
list back to a leaf in draft 09. In retrospect this was a mistake. Note that 
AFAIK buffer is currently implemented by only two vendors: Cisco and 
Alcatel-Lucent-Nokia.

The Cisco implementation has one buffer and specifies the limit as the total 
buffer size in bytes.

The Alcatel-Lucent-Nokia implementation has multiple buffers and specifies the 
limit in total messages.

If we make buffers a list, we still have three features in the model and the 
necessity for implementations that support only one buffer to augment the model 
to specify a max-elements statement. The three features are:

  feature buffer-action {
description
  "This feature indicates that the local memory logging buffer
   action is supported.";
  }

  feature buffer-limit-bytes {
description
  "This feature indicates that the local memory logging buffer
   is limited in size using a limit expressed in bytes.";
  }

  feature buffer-limit-messages {
description
  "This feature indicates that the local

Re: [netmod] syslog-model-11 single buffer vs list

2016-11-14 Thread Clyde Wildes (cwildes) 
Hi Jason,

Buffer was a subject of discussion on the netmod list most recently by Tom 
Petch who raised some questions. In an e-mail on 2016/5/6 Tom said:

“The description of log-buffer confuses me.  The buffer is circular in nature 
so there is only one of them; but it is a list keyed on 'name' so there are 
lots of them.  This leaf configures the amount number of log messages that can 
be stored in the local memory logging buffer, so there is only one of them. 
Or?”

In the same e-mail Tom also commented on the complexity of the current model:

“My comment was more on the complexity that results from having so many 
options. Other models are worse - some of the routing ones I find 
unintelligible as a result - but I raised the issue on this model because it is 
being discussed on this list where (almost) all the expertise in these matters 
resides to see if anyone else would bite.”

In a reaction to Tom’s comments, I tried to simplify by changing the buffers 
list back to a leaf in draft 09. In retrospect this was a mistake. Note that 
AFAIK buffer is currently implemented by only two vendors: Cisco and 
Alcatel-Lucent-Nokia.

The Cisco implementation has one buffer and specifies the limit as the total 
buffer size in bytes.

The Alcatel-Lucent-Nokia implementation has multiple buffers and specifies the 
limit in total messages.

If we make buffers a list, we still have three features in the model and the 
necessity for implementations that support only one buffer to augment the model 
to specify a max-elements statement. The three features are:

  feature buffer-action {
description
  "This feature indicates that the local memory logging buffer
   action is supported.";
  }

  feature buffer-limit-bytes {
description
  "This feature indicates that the local memory logging buffer
   is limited in size using a limit expressed in bytes.";
  }

  feature buffer-limit-messages {
description
  "This feature indicates that the local memory logging buffer
   is limited in size using a limit expressed in number of log
   messages.";
  }

Does it make sense to simplify the model by removing the buffer action along 
with the three features required and have vendors who implement buffer add it 
to the model through augmentation? 

A Netmod group consensus would be helpful here.

Thanks,

Clyde


On 11/13/16, 9:54 PM, "Sterne, Jason (Nokia - CA)"  
wrote:

Hi Clyde,

Somewhere in the past couple of revisions we dropped multiple memory 
buffers.  Version 8 (and a number of versions before that) had a list of 
buffers in the YANG (but it wasn't in the pyang tree).  But then version 9 
onwards seem to have a single buffer.

Can we put that back to a list ? Implementations that only support a single 
buffer can easily fit into a model that supports multiple buffers, but the 
other way around doesn't work very well.   I think it was accidently dropped 
due to some confusion over some "if-feature" comments from Tom P at one point.

(note - also add (s) to buffer to make it buffer(s) in a couple of places 
in section 3).

Regards,
Jason

-Original Message-
From: netmod [mailto:netmod-boun...@ietf.org] On Behalf Of Clyde Wildes 
(cwildes)
Sent: Monday, November 14, 2016 8:52
To: internet-dra...@ietf.org; i-d-annou...@ietf.org
Cc: netmod@ietf.org
Subject: Re: [netmod] I-D Action: draft-ietf-netmod-syslog-model-11.txt

Hi,

This draft addresses Phil Shafer’s comments and also removes references to 
TLS for now.

Thanks,

Clyde

On 11/13/16, 3:47 PM, "netmod on behalf of internet-dra...@ietf.org" 
 wrote:


A New Internet-Draft is available from the on-line Internet-Drafts 
directories.
This draft is a work item of the NETCONF Data Modeling Language of the 
IETF.

Title   : A YANG Data Model for Syslog Configuration
Authors : Clyde Wildes
  Kiran Koushik
Filename: draft-ietf-netmod-syslog-model-11.txt
Pages   : 33
Date: 2016-11-13

Abstract:
   This document describes a data model for the configuration of syslog.


The IETF datatracker status page for this draft is:
https://datatracker.ietf.org/doc/draft-ietf-netmod-syslog-model/

There's also a htmlized version available at:
https://tools.ietf.org/html/draft-ietf-netmod-syslog-model-11

A diff from the previous version is available at:
https://www.ietf.org/rfcdiff?url2=draft-ietf-netmod-syslog-model-11


Please note that it may take a couple of minutes from the time of 
submission
until the htmlized version