On Fri, Apr 18, 2014 at 07:10:16PM -0400, Susan Hares wrote:
> The assumption: 
> 
> I am assuming that the information models are not a waste of time. Jeff
> Haas' comment was isn't having information models and data models a
> duplication of effort.
> 
> First of all, RBNF and ABNF seem to be causing redundant issues in the
> draft-ietf-i2rs-rib-info-model-02 draft. 
> 
> The ABNF the delightful work in
> <http://datatracker.ietf.org/doc/draft-clarke-i2rs-traceability/>
> draft-clarke-i2rs-traceability-01 really difficult to follow.  
> 
> My assumption had been the following train would occur:
> 
> UML graphic (readable) 
>   ||
>   VV  (someday via tool) 
> UML text  (readable) 
>   ||   (via tools - I've already found some) 
>   ||
>   VV
> 
> Yang/Forces Data Model (Yang readable/Forces:Readable)   

FWIW, UML can do much of this but even some of the best authors on it note
that it can be a bit more art than science in some cases.  Since our models
to this point have been relatively simple I suspect that this train may
actually work.

> For validation of the code in the router
> 
>  Code 
> 
>   || (tools creates) 
> 
>   XML 
> 
>   ||  (tool creates) 
> 
>  Yang/Forces Data model 
> 
>   ||  (tool creates) 

I suspect this is effectively where the tool chain would traditionally stop.
You've de-marshalled the intermediate form into something that recognizes
the structured data and can dump it directly into your code.  You might
have:

> UML syntax 

UML in the middle to absorb it, but my suspicion is that a lot of existing
tools would prefer to either work on XML documents due to the related tools
like XPath.

> So.. font of wisdom from on high, can you let me know if I have
> misunderstood the point for information models and data models. 

Just to clarify my own point for the list, my concern was mostly that a
sufficiently useful data modeling language obviates much of what the info
model gives us: Structure of the information and general typing.

What something like UML gives us that isn't necessarily a property of those
data modeling languages is the inter-relationship of objects.  UML is pretty
good at that.  My only concern about graphical UML, having done a few pretty
pictures for similar info systems is that they get *big*, and splitting
those across the IETF-publishable formats may be messy.

-- Jeff

_______________________________________________
i2rs mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/i2rs

Reply via email to