really aweful naming convention I dislike

Austin Schutz tex at off.org
Tue Feb 7 10:00:46 UTC 2006


On Tue, Feb 07, 2006 at 08:18:37PM +1100, Andrew Fort wrote:
> 
> On 14/01/2006, at 3:25 AM, Justin Grote wrote:
> 
> >Indeed. The closest way I could think to coming up with a unified  
> >description for would be an XML DTD, but again, that's hard to  
> >develop a schema that could include every possible device  
> >description, and I'm not that smart :)
> 
> does NETCONF help?  (http://www.ops.ietf.org/netconf/)
> 
> one of netconf's aims is to allow vendor 'uniqueness' to be  
> expressed.  it's more about standardising the data exchange protocol  
> rather than the configuration language itself, so it doesn't solve  
> the exact problem you're talking about.  The idea is to remove the  
> messiness of expect-driven CLI sessions, which is a pretty good  
> start.  I'm sure i got part of that wrong, though :-).
> 
> for those with a bit of spare time and a cisco account, hassle your  
> account team to let you try the "Enhanced Device Interface" (their  
> netconf shell).
> 

	Once upon a time they were working on delivering config information
via snmp and XML. I found the concept quite laughable, considering the
already questionable nature of their snmp stack (Does the phrase "OID
not increasing in value" ring a bell with anyone?).
	Back before the days of massive layoff we had written a rancid->xml
converter, the idea was to eventually go the other direction as well instead
of just using templates. Now we just have auditing scripts of ever increasing
complexity examining the configs on a regular basis and skip the xml stage. It
works well enough that we've never bothered going back.
	I would recommend just trying to get the vendors to stamp out existing
bugs. *shrug* As SLAs prevent us from rotating out code with minor bugs,
often we're stuck dealing with buggy ssh servers and the like for months
at a time...

	Austin



More information about the Rancid-discuss mailing list