Re: [netlmm] Issue: Timestamp vs Sequence Number based logic

[Alexandru Petrescu <[e-mail]>]

Julien Laganier wrote:
> On Tuesday 04 September 2007, Alexandru Petrescu wrote:
>> Julien Laganier wrote:
>>> On Tuesday 04 September 2007, Alexandru Petrescu wrote:
>>>> I agree, I think SeND format (48/16bit seconds since  1970) is
>>>>  better than NTP format (32/32bit seconds since 1900) at least
>>>>  because more compatible with some Unices and because delays
>>>> the roll-over flag day to much later than 2036.  But I think
>>>> the roll-over issue is still there. IMHO SeND could get rid  of
>>>> the timestamp as well.
>>>> 
>>>> At minimal, IMHO we could write in the doc the roll-over 
>>>> P-MIPv6 year, by the Gregorian calendar (anyone cares to 
>>>> compute the year?) when P-MIPv6 will have issues and how it 
>>>> should seamlessly be dealt with it without disrupting service 
>>>> to MN.
>>> Alex, as I wrote earlier with 48bits seconds we are safe for 
>>> nearly 9 millions years.
>> Thanks, this is good to know (and sorry I have just seen now your 
>> previous message).
>> 
>> We should then write down what happens at year 
>> 9million+1picosecond: (1) how is the LMA making sure it survives 
>> the overflowing P-BU without believing it's a very old P-BU; and 
>> (2) without disrupting service to MN.
>> 
>> Don't you think so?
> 
> Maybe I'm wrong, but I somehow feel a timestamp wrap around is not a 
> problem when it is going to happen 9 millions years from now; that 
> gives us ample time (9 million years I guess) to specify an extended 
> timestamp option that works after 9 million years and takes 
> precedence over the current timestamp option. Further, when the wrap 
> around occurs we'll probably not be using IPv6 anymore...
> 
> If that is really a concern to you, would you be fine with a 128 bits
>  timestamp, that would work until the end of the universe and time?

Julien, thank you for asking.  After discussing face-to-face with
colleagues around this issue, I've been also told that in 9million years
we have enough time to conquer the galaxies and/or disappear, etc, etc.

I would like to say that you are right.

It's just that when trying to specify the usage of time the things are
very complex to specify and to implement too.

Think for example that if we use timestamps that expire in 9million
years (or 128bit at the end of universe) then the common use of time
within that timestamp is going to be very localized.  Ie much of the
bits in these timestamps don't change at all, they're the same for a
long time.  This means, grosso modo, that we use 64bit to waste bandwidth.

Alex
>> a 128 bits timestamp, that would work until the end of the universe
>>  and time
PS:128bits survives the time itself?


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

_______________________________________________
netlmm mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/netlmm