Erik Romijn wrote:
On Sat, Feb 10, 2007 at 02:06:01AM -0500, Larry Blunk wrote:
Here's my proposed format for the TABLE_DUMP_V2 type for comment.
Seeing your numbers on the RV dumps, it seems sensible to add this
index. As I hope to increase the amount of routes per peer for RIS too,
I suppose it will help us too in disk space savings.
Note that in my calculations, I didn't check for any data saving due to
saving the prefix fields only once per prefix.
So I support this format, and will be implementing it for RIS.
1 INDEX_TABLE
A minor esthetical idea: wouldn't PEER_INDEX_TABLE be clearer,
especially for future changes, if any?
Okay, that sounds better.
2 IPV4_UNICAST
One possibly stupid, certainly not all too relevant, question: why is
this called IPV4_UNICAST, while it also stores IPv6 entries?
IPV6 RIB entries would be stored in the MULTIPROTOCOL subtype.
The IPV4_UNICAST subtype is only for IPv4 Unicast RIB entries,
so there is no need to specify AFI and SAFI values (basically
saving you the 3 bytes needed to specify these).
I suppose one could use a single "RIB_ENTRY" subtype instead of separate
IPV4_UNICAST and MULTIPROTOCOL subtypes and always
include the AFI/SAFI fields.
-Larry
_______________________________________________
GROW mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/grow
