[rohc] ROHCv2 draft update - summary of changes

["Kristofer Sandlund" <kristofer.sandlund at ericsson.com>]

Hi all,

as you have already seen, we have posted a new revision of the
v2 profiles draft
(http://www.ietf.org/internet-drafts/draft-ietf-rohc-rfc3095bis-rohcv2-p
rofiles-02.txt)

As usual, below is a list of all the changes between rev -01 and -02.
Most of these have been discussed on-list and the rest are very
minor, so this list is meant to help you review the draft.

We believe that the draft now adresses all the issues brought up and
the results reflect WG consensus. What the draft needs now is
basically reviews and we (the authors) think that the best way
to trigger more reviews is to take the draft to WGLC as soon as
possible.
So, now would be a great time for you to review the draft, and also
I think it is close enough to be finished so that it should be
a good time to implement the entire thing. Any changes after this
are likely to be minor.

Changes (both text and FN changes):
********************
- MSN field in feedback renamed to Acknowledgment Number in order
to make things easier to understand
- control_crc3 (used in co_common and co_repair) is updated w.r.t.
coverage
- Timer-based compression added (time_stride, CLOCK_RESOLUTION
feedback, new encoding method). Note that due to slightly different
requirements, the CLOCK_RESOLUTION has a different name and
option number than in 3095 so that they cannot be confused.
- IC state in the decompressor renamed to RC state (repair). State
machine heavily updated.
- "differences from 3095" Appendix moved to small section in intro
part of document
- IR-PD removed.
- ts_stride defaults to 160 instead of 1
- Encoding of {ts,time}_stride in dynamic changed to 
default_or_sdvl which means that if it is not present, it reverts 
to default value. Makes repair more robust at a minimal overhead
cost.
- Updated pt_0_crc7 and pt1* for RTP-based profiles in order to
allow one more ts bit in pt1 (changed discriminators and field
lengths)
- CONTROL_FIELDS_UPDATE feedback option removed again, it is
redundant with the updates to co_common and IR above
- Unknown feedback option behavior changed, now unknown options
cause the entire feedback element to be ignored. There's a certain
risk with the previous behavior of just ignoring unknown options
- Appendix A (classification) re-written to get rid of lots of BS
text from 3095. Also added field behavior for the extension headers
for completeness
- New Appendix B for compressor implementation guidelines (very much
non-normative!) which describes some recommendations for how to
handle W-LSB, timer-based LSB on the compressor side and a bit
about context management
- res_bits in AH changed compression technique (uncompressed_value)
- auth_data in AH renamed to icv to fit with AH RFC
- Moved reorder_ratio in some co_common/repair to get the correct
octet alignment
- Moved around MSN, IP-ID and Sequence number in common/repair of
some profiles for simpler implementation parsing
- IP-ID offset encoding expressed differently (uses global control
field). No technical change, but closes bug in previous code
- Sequence Number encoding changed in co_common to use reordering
ratio and less SDVL values
- co_repair revised so that it contains a dynamic chain and is
therefore defined in "box notation"
- Bugfix in UDP-lite coverage behavior for irregular chain (this also
modified some of the flag byte encodings in co_common). Removed
all repair chains and repair_flag arguments.
- UDP checksum and ipid_offset behavior made into control fields
- Added "df" and "ip_id" as uncompressed zero-length fields in the
baseheaders encodings using ipv6 in order to not leave undefined
fields hanging loose for some formats (no change in formats, just
a FN construct for language safety)
- Bugfixed default encodings on some checksums
- Improved naming of some encoding methods for consistency (so that
"optional" has the same meaning throughout the draft)
- The ESP profile always transmits SN instead of MSN, while the MSN
is calculated locally in both compressor and decompressor (instead
of the other way around in previous revisions). This 
difference is needed because of the field length (32-bit) which
otherwise causes some unexpected 16-bit wraparound problems.
- A bunch of editorial changes with how text was expressed etc, but
that do not change anything with how things work.

BR,
  Kristofer & Ghyslain

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