I was mainly objecting to your initial
comment that no vendors support MPE Section packing, while in fact all
I have come across do support it. Now we are saying the same thing, namely
that they do not support MPE IP fragmentation. Hence we can probably stop
this discussion now.
--Tor
"William StanisLaus"
<william@erg.abdn.ac.uk> Sent by: owner-ipdvb@erg.abdn.ac.uk
02.05.2005 14:01
Please respond to
ipdvb@erg.abdn.ac.uk
To
<ipdvb@erg.abdn.ac.uk>
cc
Subject
RE: MPE Question
Dear Tor Brekke,
Sorry if I have confused, may
be I should make it more detailed J
In DSM-CC (MPE) we have section-length
to specify the size of the payload it carries, this can be maximum of 4096
bytes. Hence any IP packet as a potential payload for DSM-CC can be maximum
of 4096 bytes, this length normally discovered using L2 MTU size and IP
packets gets fragmented if the packet is greater than 4096 bytes. But theoretically,
DSM-CC supports multiple segments/fragments by itself (section number and
last section number) which gives room for bigger ip packet even greater
than 4096 bytes. Does any of the vendor uses this functionally for ip data
transmission is a question !!!. I myself never come across such support
for ip data transmission. Next, if a single DSM-CC (MPE) can carry more
than one IP packet??, I don’t understand if it is possible, since we don’t
have payload start indicator and pointer to perform IP datagram section
packing in single DSM-CC (MPE) payload, which can be decapsulated at the
other end.
Next, MPEG2-TS, for any MPEG2-TS
packet it MUST be of length 188 bytes. Any MPE (DSM-CC) packet less than
183 bytes (considering after MPEG2-TS header and payload pointer) needs
to be stuffed/padding, instead of doing so we can do section packing, by
sending next MPE packet into the same MPEG2-Transport Stream.
Hope now it is clear
Update me if I’m wrong J
-William.
From: owner-ipdvb@erg.abdn.ac.uk
[mailto:owner-ipdvb@erg.abdn.ac.uk] On Behalf Of Tor Brekke
Sent: Monday, May 02, 2005 11:25 AM
To: ipdvb@erg.abdn.ac.uk
Subject: RE: MPE Question
Hi William,
I think you are somewhat confused in the use of terminology here. Section
packing is the process of having more than one (partial) IP packet in each
MPG frame. This has nothing to do with maximum section size. All IP encapsulators
probably support MPE Section packing.
MPE Does additionally support fragmentation of an IP packet into multiple
DSM-CC sections (I guess this is what you are really talking about). This
is only useful for IP packets which do not fit into the maximum section
size of 4096 octets. This function, while defined by the MPE standard,
is not (normally?) supported by the IP encapsulators. I have to admit that
I have not checked details on ever vendor here, but have not seen this
in any of the ones I have used myself.
Regards
Tor Brekke
"William StanisLaus"
<william@erg.abdn.ac.uk>
Sent by: owner-ipdvb@erg.abdn.ac.uk
29.04.2005 17:56
Please respond to
ipdvb@erg.abdn.ac.uk
To
<ipdvb@erg.abdn.ac.uk>
cc
Subject
RE: MPE Question
Hi Siva and Bernhard,
DSM-CC which is a potential MPE of our present day DOES NOT support section
packing for normal IP data transmissions. In DSM-CC we have section length
of 12 bits, which can hold maximum of 4096 Bytes of payload data (IP in
our
discussion). In Normal scenario we don't even use this big, since section
packing is not supported practically (theoretically YES), we use DVB router
one interface as Ethernet and other as DVB, on Ethernet wire we get only
1500 bytes the max, though there is a room for section
packing(theoretically) real time packet routing don't wait for another
IP
packet, just push each and every ip packet received with MPE (DSM-CC) and
to
MPEG2-TS (supports section packing, based on the timer - reason MPEG2-TS
MUST be 188 bytes, instead of wasting bytes with stuffing we have section
packing).
When we are talking about stuffing bytes, that was perfect by Bernhard,
we
use for packet alignment, to be more specific, real time operating system
like PSOS, requires it be WORD aligned ( Multiples of Four) before
encryption algorithm to be applied for scrambling.
-William.
> -----Original Message-----
> From: owner-ipdvb@erg.abdn.ac.uk [mailto:owner-ipdvb@erg.abdn.ac.uk]
On
> Behalf Of Bernhard Collini-Nocker
> Sent: Friday, April 29, 2005 6:55 AM
> To: ipdvb@erg.abdn.ac.uk
> Subject: Re: MPE Question
>
> Dear Siva,
>
> you are right in that I had understood your question different.
>
> Generally Sections do allow for numbering that is, one can put a payload
> (for example a DSM-CC object) into a number of Sections and with
> section_number and last_section_number the reassebler knows what to
do.
> MPE uses this Section mechanism, so multiple Section operations should
> be possible. Actually I have never tested whether MPE decapsulators
do
> support this in real, in principle they should (again the DVB
> databroadacst standard does not say too much, if anything, about it)
but
> in DSM-CC Obejct carousels is a very usual mode of operation.
>
> Siva Veerepalli wrote:
> > Thanks for the response Bernhard, but I am a bit
> > confused about the answer. Please see my question
> > below:
> >
> > --- Bernhard Collini-Nocker <bnocker@cosy.sbg.ac.at>
> > wrote:
> >
> >>Siva Veerepalli wrote:
> [...]
> > I understand that two MPE sections could probably
> > start in the same TS packet (I guess that is what you
> > mean by section packing).
>
> Yes.
>
> > My original questions was, what is the usual practice
> > with regards to fragmenting an IP datagram over
> > multiple MPE sections (irrespective of how these
> > sections are transported in TS packets)? The DVB-H
> > (data broadcast standard) allows for this, however,
> > the IPDC interim specification requires that one
> > datagram be encapsulated entirely in one mpe section
> > i.e., no datagram fragmentation over multiple MPE
> > sections.
>
> Now the confusion starts when one introduces the wording of "over
> multiple MPE sections", because there is no such thing as a MPE
section
> fragmentation.
> I will have to read the DVB-H standard again in order to find out
what
> it really says. I guess the only real issues the DVB-H wrt to IP is
> addressing is the necessity of a IP datagram encapsulated in MPE is
to
> support time slicing (ie a burst transmission of the resulting TS
> packets to allow for power saving) and MPE-FEC (adding redundant code
to
> allow for reconstruction). In that case I woould assume that it is
very
> reasonable (especially with legacy equipment) to avoid all potential
> interpretations of MPE wrt multiple section operation and section
packing.
>
> > thanks,
> > Siva
> [...]
>
> Yor are very welcome,
> and I hope I got THE expected answer closer,
> Bernhard