On Fri, Nov 25, 2005 at 12:35:27PM +0100, Mohacsi Janos wrote:
>
>
>
>
> On Fri, 25 Nov 2005, Stig Venaas wrote:
>
> >On Fri, Nov 25, 2005 at 12:11:18PM +0100, Mohacsi Janos wrote:
> >>
> >>
> >>
> >>
> >>On Fri, 25 Nov 2005, Vishwas Manral wrote:
> >>
> >>>Hi Mohacsi,
> >>>
> >>>LWAPP encapsulation, IPv6-in-IPv6 etc.
> >>
> >>I have to study LWAPP encapsulation - currently I have no opinion. In the
> >>IPv6-in-IPv6 encapsulation It is completely possible that tunnel endpoint
> >>has to fragment the packet. Multiple level encapsulation might cause
> >>multiple level of fragmentation. This is the price of if encapsulation....
> >
> >Right, so I think a typical implementation would as you say, let each
> >fragment have the Path MTU size (at least 1280), while the last might
> >be smaller.
> >
> >But assume you have say a 1500 byte packet. Instead of fragmenting so
> >that the first fragment has say 1280 bytes, why not split packet in
> >two equal sized fragments. That way you avoid further fragmentation
> >which might occur if there is further encapsulation taking place later
> >on the path.
> >
> >BTW, I know that fragmentation order might vary, e.g. Linux at times
> >send (or used to send) fragments in reverse order.
> >
>
> You halving approach not very useful for 9K packets, which is tend to be
> common on GE and 10GE.
>
The sender could when doing fragmentation make an effort to send the smallest
amount of fragments and distribute the payload as equally as possible across
the fragments. Would that improve things?
Best regards
--
Programmer
Edgar E. Iglesias <[EMAIL PROTECTED]> 46.46.272.1946
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
