Martin Bjorklund wrote:
Andy Bierman <ietf@andybierman.com> wrote:
Martin Bjorklund wrote:
....
I disagree.
This is in there by design to support configuration servers.
For example:
<edit-config>
<target><running/></target>
<default-operation>none</default-operation>
<test-option>test-then-set</test-option>
<error-option>rollback-on-error</error-option>
<url>https://example.com/config/myConfig.xml</url>
</edit-config>
I agree with you that this would be useful. But the text in 8.8.5.1
says that the url should specfiy a local file. Maybe that sentence
just should be removed? I don't quite see the point of that
restriction.
To be picky, it says should instead of SHOULD.
It really doesn't mean anything normative.
Yes I noticed that. Ok, so the intention is that the url can be
remote in this case. This makes perfect sense to me.
Phil's reply was that <validate> needs a complete configuration. Are
you saying that it doesn't? The text says "... the <config> element
containing the configuration subtree to validate."
I don't think the text says that anywhere.
I should have mentioned the reference, sorry, it's in 8.6.4.1. (My
point here was that the term "configuration subtree" can be
interpreted as a partial config.)
Certainly, the inline config XML does not have to be a complete
configuration.
Ok. But then the validation doesn't give you much more than syntactic
checks. If you first validate a partial config and then try to apply
it with edit-config, the resulting configuration might be invalid.
It depends on the data model!
In my models, there is a conceptual difference between
parameter sets and mere complexTypes (unlike XSD BTW).
A parameter set can be validated for referential integrity, but
not external dependencies.
This is more than syntax-check but less than full-config validate.
BTW, it doesn't say anywhere in the spec that the agent guarantees
the input that passed a <validate> will always work in a subsequent
<edit-config> or <commit>. It is a best-effort feature.
So if you do <validate> on the candidate, it will validate the full
config.
Yes
And <validate> on inline config will validate a partial
config (e.g. it must allow incomplete relations).
Yes
Then the question
is what should <validate> on a local file do? Should it treat it as a
full config or partial?
IMO, it doesn't really matter where the input comes from.
The agent can validate what it gets against what it has, the best it can.
To repeat myself -- it doesn't really say anywhere in the spec
that URL means MUST be a complete config file, or MUST be
same or different format than the representation of the built-in
configuration datastores.
The <url> feature is just supposed to be input redirection (in this case).
/martin
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/>