updated ipv4 id specification
DESCRIPTION
Updated IPv4 ID Specification. IETF 77 Joe Touch, USC/ISI. Quick review. Current hosts don’t ensure ID uniqueness within 2MSL Recognize existing practice Explicitly limit ID to fragmentation Align IPv4 and IPv6 ID and fragmentation. -03 / intarea mods. Issued as inarea doc Text mods - PowerPoint PPT PresentationTRANSCRIPT
04/20/23 19:41 1
Updated IPv4 ID Specification
IETF 77
Joe Touch, USC/ISI
Quick review
Current hosts don’t ensure ID uniqueness within 2MSL Recognize existing practice
Explicitly limit ID to fragmentation Align IPv4 and IPv6 ID and
fragmentation
04/20/23 19:41 2
-03 / intarea mods
Issued as inarea doc Text mods
Reformatted pseudocode fragments Clarified requirements text Mathis ack’d only
04/20/23 19:41 3
IPv4 vs. IPv6 IDs
IPv4 IPv6
Length 16 bits 32 bits
Presence All packets Only fragments
Frag. Location Anywhere “Source” only
Uses Frag/reassyDiscard dups
Frag/reassy
BW @1.5K 6.4 Mbps 19 Tbps
BW @64K 280 Mbps 820 Tbps
04/20/23 19:41 4
ID MSL uniqueness already ignored
Summary of changes
SHOULD limit fragmentation in IPv4 Only at source (DF=1) Only where use is safe (ID unique or integrity
check)
MUST NOT fragment if ID not unique MUST NOT reuse ID if fragmented/able
MUST NOT use for duplicate detection MUST ignore overlap fragments
Now possible only by net duplication as transient
04/20/23 19:41 5
Remaining issues
Clarify ‘maximum jitter’ “time during which reordering is expected &
tolerated.”
Clarify IPv6 examples IPv6 ID also in PTB ICMP, only for v6 talking to v4 May update RFC 2460 (?)
Clarify why overlaps are prohibited Never retransmit with same ID Now impacts only in-net replication (transient
impact)
04/20/23 19:41 6
More issues
Impact on NATs NATs act as hosts; SHOULD set DF, not
fragment
Note that there are boxes that ignore DF With predictable results ;-)
Transition plan Do we need one? – it might help with ignore-
DF case When to cease?
04/20/23 19:41 7