Re: [Idr] draft-snijders-idr-rfc8203bis

[Job Snijders <[e-mail]>]

On Thu, Dec 14, 2017 at 12:43:29PM -0500, Jeffrey Haas wrote:
> > On Thu, Dec 14, 2017 at 05:08:00PM +0000, [e-mail] wrote:
> >>> From: Idr [e-mail] On Behalf Of Peter Hessler
> >>> The len field is 8 bits, which allows for a max of 255.
> >>> 
> >>> If longer is desired, then we'll need a new error code point.
> >> 
> >> Alternatively, another field may be added after the existing
> >> "Shutdown Communication" field.  Either as a replacement (setting
> >> the length of the first field to zero) or as a concatenation. 
> > 
> > Bruno is right (of course, since the length field was his idea to
> > begin with :-), there are options available to extend beyond 255 if
> > needed.
> 
> If the limit was 127 rather than 128, the top bit could have been
> used. 
> 
> The closest thing that can be done will depend partly on what deployed
> implementations do when 128 is exceed.  The spec says nothing right
> now... and thus you might not do anything beyond saying "up to 255 is
> fine".  

Most of the open source implementations I'm aware of will log a message
"invalid shutdown communication", and perhaps print a hexdump of that
data. Since we're relatively early in the deployment cycle, I think it
is feasible to upgrade the field and patch the existing implementations.

> To get beyond 255, the simplest but gross one is to say "there may be
> a second message after the shutdown communication field".

If (and only if) we pick the route of going beyond 255, I'd consider
deprecating RFC 8203 and have the new thing require the first byte to be
a zero. I don't think there should be two 'shutdown communications'.

> -- Jeff (and people wonder why I'm pushy about using large "length"
> fields.)

Not anymore! :)

Kind regards,

Job