Re: [manet] New definition of term 'TLV' in PacketBB

[Alexandru Petrescu <alexandru.petrescu at gmail.com>]

Thomas Clausen wrote:
> Alex,
>
> I appreciate your reviewing of the MANET WG specs.
>
> Generally, I think that it might be advantageous if you try to 
> understand WHY the specs are the way they are, rather than to launch
>  into observing a difference with some other similar mechanism and
> then taking this as an indication that something is broken.

Sorry, I didn't mean something was broken; just that it may be 
unnecessarily complex.

> TLVs, for example, are TLVs, with the addendum of specifying "to
> which entities they apply". This is necessary since it allows doing
> efficient address aggregation. The alternative -- naive TLVs without
> such information -- would lead to a much less efficient address
> aggregation. (Notice that this is a simplified description, but that
> it should get the point through)

The other alternative of overspecifying new ways of encoding addresses
risks obviously to create large packets for simpler cases.  Whereas I
agree head-mid-tail encodings of addresses can lead to smaller packets
when several similar addresses are sent, apparently it sends more bytes
than 'naive' TLV when only one address is sent.   And this may be less
efficient for MANEMO where often only one address/prefix is needed (the
Mobile Network Prefix).

> Since one of the MANET wg's design-dogmas is "efficient when
> sensible", many such trade-offs have been considered and discussed
> over the years, and packetbb reflects the result of a number of such
> discussions.

I agree with general consensus.  I will send a separate email trying to
identify how PacketBB can encode an IPv6 Mobile Network Prefix and then
compare that encoding with the other existing ways of encoding Mobile
Network Prefixes.

Alex

> Thomas
>
> On Jun 19, 2007, at 4:11 PM, Alexandru Petrescu wrote:
>
> > Dearlove, Christopher (UK) wrote:
> > > > I think the term TLV (type-length-value) is overloaded by the 
> > > > PacketBB
> > > spec: There are many uses of TLV in many places. Some are simple,
> > > some are more complicated. This one does the job it's intended to
> > > do.
> >
> > But it's different than what rfc2460 calls a 'TLV'.
> >
> > > You actually can't do that job with just a type, a length and a
> > > value as you then couldn't apply information to just some of the
> > > addresses in an address block. And that's essential. I'm afraid I
> > > don't find this particular criticism constructive, so I'm not
> > > going to discuss this point further.
> >
> > Ok, I again went into MANET document debugging instead of
> > contending just reading them :-( Generally speaking, I don't think
> > it's constructive to ignore existence of already defined TLV
> > mechanisms, address encoding mechanisms and code availability.
> >
> > Alex
> >
> >
> > ______________________________________________________________________
> >  This email has been scanned by the MessageLabs Email Security
> > System. For more information please visit
> > http://www.messagelabs.com/email 
> > ______________________________________________________________________
> >
> >
> >
> > _______________________________________________ manet mailing list 
> > manet at ietf.org https://www1.ietf.org/mailman/listinfo/manet


______________________________________________________________________
This email has been scanned by the MessageLabs Email Security System.
For more information please visit http://www.messagelabs.com/email 
______________________________________________________________________


_______________________________________________
manet mailing list
manet at ietf.org
https://www1.ietf.org/mailman/listinfo/manet