Re: <validate> operation somewhat broken

[Andy Bierman <ietf@andybierman.com>]

Andy Bierman wrote:
> Hi,
>
> IMO, the <validate> command is not actually useful for 'inline'
> configuration validation.

I think the WG should decide what to do about this problem:

   1) Ignore it
   2) Document it
   3) Fix it

IMO, the proper engineering fix requires 2 independent changes:

 1) Add a 'test-only' enumeration to the (edit-config) test-option
    parameter.  (Remember that this entire parameter is only supported
    if the :validate capability is supported. This change could
    be done as a validate-2.0 capability, and not change the 1.0 capability.


 2) Remove configInlineType as one of the options for the 'source'
    parameter in the validate operation. (Make it the same data type
    as the getConfig source parameter).  The normative text only
    mentions 'config name' as a possible value for the 'source' parameter,
    so perhaps the URL support in the XSD is wrong as well.

    The validate command is only used to check an entire configuration,
    so configInlineType is wrong, even if no other change is made.
    The 'operation' attribute should not be present in the data nodes,
    for the validate operation, unlike the edit-config operation.

I remember we had a test-only mode for edit-config and took it out because
the validate operation covered it.  Maybe this is not important
enough to people that we fix it.  We will see.


Andy




> The reason the edit-config operation does not have a 'test-only' mode
> is because the validate operation is supposed to provide that feature.
> Except it doesn't actually work.
>
> Problem 1) Assumed target.  Since the <source> parameter is a choice,
> the manager cannot specify a target -- it is assumed to be either
> the <running> or <candidate> config.  There is no way to specify
> a user-named config (url) as the target if the <config> inline mode is 
> used.
>
> Problem 2) Assumed default operation.  There is no way to pass the
> default-operation parameter that would be used in the <edit-config>
> along with the <config> element.  The <validate> operation is supposed
> to come up with the same answer as an <edit-config> would in this mode,
> but without this critical parameter, that is impossible.
>
>
> The text in 8.6.4.1 does not match the XSD.
> It does not indicate that 'inline config' or 'url' modes
> are allowed here.  The only example shows the 'config name' variant.
> Why didn't anybody bring this up until now?  Does anybody support
> this operation?
>
>
> 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/>