Re: [hybi] Framing Take VI (a compromise proposal)

John Tamplin wrote:
> On Tue, Aug 17, 2010 at 7:50 PM, Thomson, Martin 
> <Martin.Thomson@andrew.com <mailto:Martin.Thomson@andrew.com>> wrote:
>
>     That wasn’t the suggestion at all.  Binary/text was included.  The
>     proposal was length + opcode.  Two opcodes only…for now.
>
>      
>
>     The other stuff (fragmentation, channels, etc) are well enough
>     understood now to say that they can be built in later. Hence, we
>     reserve the bits and plan to build them later (or not at all). 
>
>
> I think we do know that we have a use-case for fragmentation, as it 
> has been accepted as a requirement for this protocol.  As previously 
> mentioned, the inclusion of a total message length was required to 
> remove objections to fragmentation.  So, other than having two options 
> for the frame length to preserve efficient encoding of small messages, 
> I think the current proposal is about as minimal as it can get.

A variable-length frame-header is not as minimal as it can get.

In your design, initial fragments have a 18 byte header (!), 
non-fragmented messages > 127 have 10 byte header, non-initial fragments 
are 10 bytes, and 127 byte messages (shorter than a tweet) have a 
special 2 byte header.

That is not a minimal proposal.

-- Scott

>
> I think if we leave out fragmentation and planned extensibility from 
> the base protocol, we haven't really accomplished anything and we will 
> either wind up having to make incompatible changes later or having 
> suboptimal framing once we do define some extensions.
>
> Let's say we did strip out fragmentation, and even defined the INI/FIN 
> bits as proposed, but required that they be 1 in the initial version. 
>  Then, any intermediaries and receivers that are written now will 
> break when fragmentation is added later, and experience shows that 
> those intermediaries will be very slowly upgraded.  I think that would 
> add years to when fragmentation could usefully be used, so I think it 
> is entirely appropriate that fragmentation support be in the base 
> protocol.
>
> -- 
> John A. Tamplin
> Software Engineer (GWT), Google
> ------------------------------------------------------------------------
>
> _______________________________________________
> hybi mailing list
> hybi@ietf.org
> https://www.ietf.org/mailman/listinfo/hybi
>   