RE: Proposal for new last call issue: Some unprocessed headers should stay

>if the semantics of the header are such that it can be applied 
>on the back link breaking the application if it was not also 
>applied on the previous node on its forward link, what 
>prevents the application from breaking at B? Unless, of 
>course, B and C can understand a header differently, which 
>would be insane. 8-)

The application wouldn't break at B as C wouldn't see the feature and
attempt to apply it to the B<--C link.

>As for the store and forward semantics, why doesn't the node 
>that knows it's before a weak link play a special role, like 
>"weakLinkProxy"? I would then probably want to address all 
>weakLinkProxies to store and forward anyway.

That won't change anything--remember that we are not talking about just
flipping the default for "next" but for all roles. That is, if I target
a header block at the role "weakLinkProxy" and it didn't understand that
header block then flipping the default would still forward the header
block which could cause breakage further down the line (B<-->C link).

Henrik

Received on Wednesday, 16 October 2002 17:26:36 UTC