-
From: Gatot Susilo <gatot.susilo@alcatel.com>
-
Date: Thu, 17 Feb 2005 15:27:45 -0500
-
CC: mpls@lists.ietf.org
-
Organization: Alcatel Canada
Hi Kireeti and George:
I have some comments for the draft.
1/ Timestamp on echo request message.
It's expected that the source and destination node will timestamp the
echo message. It would be expected that each node will be synchronized
by NTP server. If I am not mistaken, the NTP server will give accuracy
about 10msec. Therefore, between two nodes synchronized by the same NTP
server may have time difference up to 20msec. On the broadband network,
the round-trip-time of echo message may be less than 10msec.
This may bring an issue that round-trip-time will be less than the
trip-time.
Also when the destination node may have time lag (i.e., +/- 10msec)
w.r.t the source node, the trip time will be a negative value (i.e. rx.
timestamp - tx. timestamp).
Could you please explain what the benefit of having timestamp on the
echo message and how do we deal with the issue above ?
If there is no benefit, probably it should be removed.
2/ Mandatory/Optional TLV.
The mandatory/optional TLV is identified by high-order bit. If the
high-order bit is not set, the TLV is mandatory, otherwise it's optional.
To clarify this, would it be possible to include this "high-order"
bit in the TLV format diagram?
Hence, the TLV format will look like this:
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|F| Type | Length |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Value |
. .
. .
. .
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
F - set to 0 for mandatory TLV and set to 1 for optional TLV.
3/ Errored TLV:
The draft stated that the Errored TLV has type=32678 (i.e., 0x7FA6).
As mentioned previously that the optional TLV sets the high-order bit ,
should this Errored TLV have value=0xFFA6 (i.e. 65446). This can also be
fixed with the proposed TLV format above.
4/ Typo on length of BGP Labelled IPv4 FEC and CR-LDP FEC.
The length of BGP Labelled IPv4 FEC should be "9" instead of "10".
The length of CR-LDP FEC should be "8" instead of "6".
Cheers,
/Gatot