Re: [netmod] Alissa Cooper's Abstain on draft-ietf-netmod-artwork-folding-09: (with COMMENT)

[Martin Bjorklund <[e-mail]>]

Erik Auerswald <[e-mail]> wrote:
> Hi Martin,
> 
> On Thu, Sep 05, 2019 at 05:17:30PM +0200, Martin Bjorklund wrote:
> > Kent Watsen <[e-mail]> wrote:
> > > 
> > > >> There has been discussion about how embedding YANG models in RFCs seems like a
> > > >> poor fit for a number of reasons. By standardizing line-folding mechanisms and
> > > >> claiming them as a best practice, this document reinforces the root of that
> > > >> problem rather than trying to fix it.
> > > > 
> > > > Well said, I agree with Alissa's conclusion.
> > 
> > But the algorithm in the document isn't supposed to be used for YANG
> > modules.  It is supposed to be used primarily for XML and JSON
> > snippets (etc).
> 
> Technically, XML and JSON are whitespace-agnostic

Whitespaces are significant in XML element values, which means that we
have to write e.g.

    <discontinuity-time>2013-04-01T03:00:00+00:00</discontinuity-time>

With indentation for readability we quickly run out of horizontal
space.

> and often can be
> "folded" manually without any folding indicators.  That seems to be
> true for YANG as well.

It is true for YANG, which is why I wrote that this algorithm isn't
supposed to be used for YANG.  In fact, the document says:

   It is RECOMMENDED that authors do as much as possible within the
   selected format to avoid long lines.

> Even Python code usually can be kept under any
> reasonable line length.
> 
> Very long names or values could require algorithmic folding, i.e.,
> a generic line-folding mechanism.
> 
> I personally would try to keep the line length of any code inside an
> RFC short enough to not need additional algorithmic folding.  But if
> this does not work, e.g., because some name or value is just too long,
> having a recommended algorithm seems to be better than every author
> making up another ad-hoc folding scheme.
> 
> The I-D.ietf-netmod-artwork-folding seems to try to provide a recommended
> algorithm for the case where a generic line-folding mechanism is needed.

Exactly!


/martin


> 
> Overly long lines sometimes result in lost information in HTML or PDF
> documents, if a complex layout is used, but no mechanism for dealing with
> unexpectedly long lines is included.  I have seen too many examples of
> this to believe that this cannot happen with RFCs.
> 
> Thanks,
> Erik
>