Guys,
I would have a
small observation on the comment number #112 . Although I assume that you
guys had a technical review of all these comments I should point out some
details.
Comment # 112 proposes a method of errored PDU's recovery in case
of errored PDU header. In this particular situation as you loose the length
information if no specific recovery is done , all PDU's contained in a burst
can be lost.
The
problem is that aside of introducing a new optional mode that is changing
the frame (burst) structure ,it does not add value on the technical
side for few reasons:
1.
There are methods for next header recovery that are widely used in ATM for
instance, without any OVERHEAD addition and structure change. Some fields
in the header have some possible patterns which any correlator can
identify .Starting from there, any simple HW machine can position on the
header 'candidate' .When the header candidate is localized the simplest thing
is to check the CRC at the end of it. If it matches that means the header is
identified. The probability for mistake is almost null.
There are two conditions that have to be met:
- the pattern
- CRC on the candidate header
sequence
This
algorithm does not require any change ,any overhead and any
option.
2.
The method presented in the contribution is not good , aside added overhead and special
framing. These pointers present in the sequence are exposed to errors as
much as any bit in the burst. Then probability of that pointer being errored
is the same. That errored pointer does not have any protection and any
indication that is errored or not. Based on that you can get in a worse
situation ... The only probability on this method to work is
to be lucky enough in a noisy environment to receive correct
pointers.
3.
We talked about this item in the last Orlando meeting and as I remember was
rejected on technical ground. I replied to the comment in recirculation but I
did not get any feedback .
Cheers,
Radu.