On Tue, Feb 15, 2022 at 04:45:28PM +0100, Juergen Gross wrote:
> On 15.02.22 16:42, Anthony PERARD wrote:
> > On Thu, Feb 10, 2022 at 12:16:20PM +0100, Juergen Gross wrote:
> > > +The "Connection error indicator" is used to let the server indicate it 
> > > has
> > > +detected some error that led to deactivation of the connection by the 
> > > server.
> > > +If the feature has been advertised then the "Connection error indicator" 
> > > may
> > > +take the following values:
> > > +
> > > +Value   Description
> > > +-----------------------------------------------------------------
> > > +0       No error, connection is valid
> > > +1       Communication problems (event channel not functional)
> > > +2       Inconsistent producer or consumer offset
> > > +3       Protocol violation (client data package too long)
> > 
> > Is this meant to be the only possible error value? If in the future we
> > want to add more possible error, does it going to need a new feature
> > bit and maybe a new error field?
> 
> No, as the guest is not opting into this feature, but just gets it
> presented, there is no need to have another bit for new error values.

This probably needs to be spelled out in the documents that.

> Note that this is a purely informational interface. The error value
> (other than 0) is only for diagnostic purposes, there is no way a
> guest could react in a sane way to a specific error case.

This could also be spelled out in the document, in the new section "The
connection error feature", that a value other than 0 is a connection
error even if the error number isn't known to the client.

Thanks,

-- 
Anthony PERARD

Reply via email to