Re: [netconf] Adoption call for draft-kwatsen-netconf-http-client-server-04

----- Original Message -----
From: "Rob Wilton (rwilton)" <rwilton@cisco.com>
Sent: Monday, November 11, 2019 2:26 PM

Hi Kent,

I broadly agree with Martin here, in that I think that we want to get
all of the "Client-Server" drafts finished.

Defining groupings in one place is probably beneficial and I as such as
support the adoption of this draft, but not if it going to cause the
other drafts that depend on these modules to end up being blocked for
another 8 months, or a year.

<tp>
Well, my view of groupings and similar constructs in a variety of
languages is that they are mostly harmful, making the code more complex
and harder to understand, requiring a reader to keep having to branch
off and look up something else somewhere else.

They are beneficial only when they encompass a well-defined meaningful
entity, object, idea, collection of data.  When the boundaries are
ill-defined, I find them harmful.

What I see is the same 'n' lines of code, where 'n' is a small, even
very small, integer, appearing in more than one place and then those 'n'
lines being taken out of context, given a wrapping, a name, a separate
evolutionary path just to save <<n lines of code.  As I say, mostly
harmful.

I do agree that any further delay in the base modules that depend on
this module is to be avoided.

Tom Petch

Can we quickly find a name for these modules that is good enough for us,
and would also be acceptable to the httpbis WG?

Perhaps "http-rest" instead of "http" or "restful"?

Thanks,
Rob

From: netconf <netconf-bounces@ietf.org> On Behalf Of Kent Watsen
Sent: 08 November 2019 01:45
To: Martin Bjorklund <mbj@tail-f.com>
Cc: httpbis-chairs@ietf.org; netconf@ietf.org
Subject: Re: [netconf] Adoption call for
draft-kwatsen-netconf-http-client-server-04

Hi Martin,

I think it is quite clear that the WG doesn't want more delays.

Is it?   I only know that Balazs K. wishes to have the first three
drafts (crypto-types, truststore, and keystore) completed soon.  The
Nokia/BBF folks were interested in netconf-server, but forked long ago.
Otherwise, I'm unaware of anyone stating a preference.   For myself, I
only hope to achieve a good result.

So I think that either we simply don't add this grouping *at this
point*,
and define the necessary nodes in the higher-level modules, even
though it is not perfect; or we define the "restful" module with these
groupings, even though that is also perhaps not perfect.

ietf-restful-client and ietf-restful-server sound nice, but not so much
in a stack:

    tcp-server-parameters : { ... },
    tls-server-parameters : { ... },
    restful-server-parameters : { .... },        <---- end-user says
"what's that?"
    restconf-server-parameters : { ... }

or (for https-notif)

    tcp-client-parameters : { ... },
    tls-client-parameters : { ... },
    restful-client-parameters : { .... }    <---- end-user says "what's
that?"

Actually, the above is purposely inaccurate.   If you recall, we
switched to having container-less groupings, thus the consuming module
can name the container whatever it wants, so the net-net is no impact
other than changing the import and uses statements.   Though I still
don't know why we should do even this, given the lack of any response to
my Oct 30th message.

Kent // contributor

------------------------------------------------------------------------
--------

> _______________________________________________
> netconf mailing list
> netconf@ietf.org
> https://www.ietf.org/mailman/listinfo/netconf
>