On Fri, Dec 18, 2015 at 04:16:04PM +0100, Ladislav Lhotka wrote:
> 
> > On 18 Dec 2015, at 15:49, Juergen Schoenwaelder 
> > <[email protected]> wrote:
> > 
> > On Fri, Dec 18, 2015 at 03:22:48PM +0100, Ladislav Lhotka wrote:
> >> 
> >> Is it not? I would say it severely restricts the workflow for the data 
> >> model development. The ultra-conservative update rules essentially permit 
> >> only incremental changes to published modules. This would be fine if the 
> >> data model landscape already was reasonably stable. We are not that far 
> >> though, and everything is in flux. So I believe we would be much better 
> >> off with "release early - release often" strategy, which is made 
> >> impossible by the existing update rules.
> >> 
> > 
> > There is a "release early - release often cycle" in the IETF process -
> > it is called the Internet Drafts stage. Unfortunately, often people
> > wait for things to stabilize (becoming an RFC) before implementing. I
> 
> There are other disadvantages to I-Ds, for example that they have to be 
> updated every six months. It is actually funny: RFC used to mean "request for 
> comments", then later I-D acquired this role, so now we probably need a 
> "drafty draft" category.
> 
> > assume we would have fewer but more solid data models if they would
> > all come along with running code behind them (and ideally > 1
> > independent implementations). The problem might be "us" and not the
> > update rules.
> 
> The update rules mean that it is risky to publish a data model in an RFC. And 
> indeed, if there is a need, for one reason or another, to restructure, for 
> example, ietf-interfaces, it will have to become a new module 
> (ietf-interfaces-dash?) and the revision history will be broken. Doing this 
> more than once would turn the data model ecosystem into a mess.

No, it does not.
 
> Backward compatibility is nice, but making it an absolute requirement, which 
> is even ingrained in the data modelling language specification, is IMO absurd.
>

I can grab interface statistics from pretty much all devices on my
network using a single data model and SNMP. This is a feature, not a
bug. You may find this not worth the effort or you might even find
this absurd in todays world. I do not have to share this view.

Can we stop here and get back to the I-D discussed in this thread? If
you want to complain about absurd YANG update rules, please do so
using a separate thread.

/js

-- 
Juergen Schoenwaelder           Jacobs University Bremen gGmbH
Phone: +49 421 200 3587         Campus Ring 1 | 28759 Bremen | Germany
Fax:   +49 421 200 3103         <http://www.jacobs-university.de/>

_______________________________________________
netmod mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/netmod

Reply via email to