[Speex-dev] draft-ietf-avt-rtp-speex-01.txt

[Aymeric Moizard]

comment inline.

On Wed, 16 May 2007, Jean-Marc Valin wrote:

>> Page 3:
>>
>>    To be compliant with this specification, implementations MUST support
>>    8 kHz sampling rate (narrowband)" and SHOULD support 8 kbps bitrate.
>>    The sampling rate MUST be 8, 16 or 32 kHz.
>>
>> There is a type above after (narrowband), there is a " extra character.
>>
>> I don't understand what is the motivation to specify "SHOULD support 8
>> kbps bitrate".
>
> The main idea is that Speex supports many bit-rates, but for one reason
> or another, some modes may be left out in implementations (e.g. for RAM
> or network reasons). What we're saying here is that you should make an
> effoft to at least support (and offer) the 8 kbps mode to maximise
> compatibility.

I understood this. But as you may know: the SDP parameters are PROPOSAL
only and a remote application might use another "mode": this typically
lead to interoperability issue and you should advise in the specification
to always support all "modes". I understand this can be seen as a 
limitation, but in real world, it will not be acceptable to support
only a few mode among the provided ones.

>> Page 8:
>>
>>    Optional parameters:
>>
>>       ptime: see RFC 4566.  SHOULD be a multiple of 20 msec.
>>
>>       maxptime: see RFC 4566.  SHOULD be a multiple of 20 msec.
>>
>> In real world, many SIP application use either 20 or 30ms. This
>> ptime parameter is really not reliable for negotiation... On possible
>> way to handle non multiple would be to take the right above value:
>> if 30ms is specify, then recommand to use 40ms for speex.
>
> Actually, it needs to be "MUST" be a multiple of 20 ms because no matter
> what happens, Speex frames are 20 ms long. I expect most clients would
> use 20 ms, as it corresponds to one packet. As to what we need to do if
> the ptime is invalid, I'm not quite sure, though maybe as you say round
> it up (or down?).

I understand that speex needs multiple 20ms packets: for speex 
"packetisation interval MUST be a multiple of 20ms", but you have
to provide a specification compliant with other ones: "ptime" can
have any other value and there can't be a MUST there.

Round it up is a much better idea: usually, 30ms is used when
20ms would introduce too much bandwidth overhead: if you round
it down, then you would get less quality.

>> Page 10:
>>
>>    The value of the sampling frequency is typically 8000 for narrow band
>>    operation, 16000 for wide band operation, and 32000 for ultra-wide
>>    band operation.
>>
>> The word "typically" means to me that it could be something else than
>> 8000, 16000 or 32000: I would recommend to make it clear:
>>
>>    The value of the sampling frequency MUST be either 8000 for narrow band
>>    operation, 16000 for wide band operation, and 32000 for ultra-wide
>>    band operation.
>
> Agreed.

Good.

>>       ptime: duration of each packet in milliseconds.
>>
>> http://www.ietf.org/rfc/rfc4566.txt specify that in the ptime definition:
>> "it is intended as a recommendation for the encoding/packetisation of
>> audio". Thus, I would recommend to specify the same text as in rfc3264
>> for sdp offer/answer model:
>>
>>    "If the ptime attribute is present for a stream, it indicates the
>>    desired packetization interval that the offerer would like to
>>    receive.  The ptime attribute MUST be greater than zero."
>>
>> It might also be a good idea to say that even if an offerer would like
>> to receive 20ms, the sender MAY use a different packetization interval...
>> This is the origin of numerous interop issue with speex in SIP
>> applications.
>
> Sounds fair. Just curious, what's the exact interop issue?

Some application "allocate" a buffer based on the "ptime": thus
they copy 20ms of PCMU data each time they get a packet even
if they receive packets each 30ms...

The sound cards play 2/3 of data received... This happen more
than you would imagine. Look at this implementation of current
iLBC in asterisk:

http://www.asteriskpbx.org/doxygen/trunk/codec__ilbc_8c-source.html

The initEncode is called with a #define

#define ILBC_MS         30
/* #define ILBC_MS         20 */

static int ilbctolin_new(struct ast_trans_pvt *pvt)
{
     struct ilbc_coder_pvt *tmp = pvt->pvt;

     initDecode(&tmp->dec, ILBC_MS, USE_ILBC_ENHANCER);

     return 0;
}

The above means: If you negotiate a different packetisation interval: 
just recompile   ;(

Weird...

Old speex implementation in asterisk was doing the same, it seems it's
now fixed in latest version.

>>       sr: actual sample rate in Hz.
>>
>>       ebw: encoding bandwidth - either 'narrow' or 'wide' or 'ultra'
>>       (corresponds to nominal 8000, 16000, and 32000 Hz sampling rates).
>>
>> Both the "sr" and "ebw" conflicts with speex/XXXX rtpmap. I really
>> recommend to remote both those definition so that application will
>> configure themselves using either speex/8000, speex/16000, speex/32000.
>> Having 3 way to specify sampling rate is a nightmare for interop.
>
> Had missed that one. It definitely makes sense. The original draft
> allowed specifying using the narrowband/wideband encoder independently
> of the sampling rate, but in retrospect, that was just plain wrong.

I'm so happy of that answer: I hope there is a consensus here.

>> Page 11:
>>
>>       mode: Speex encoding mode.  Can be {1,2,3,4,5,6,any} defaults to 3
>>       in narrowband, 6 in wide and ultra-wide.
>>
>> I always asked for a "table" in the specification here providing link
>> between "mode" and "bitrate". Else, you get those mails:
>>
>> http://lists.xiph.org/pipermail/speex-dev/2006-March/004288.html
>>
>> If I get it right, the table is there:
>>    http://www.speex.org/docs/manual/speex-manual/node10.html
>>    Table 4: Quality versus bit-rate
>>
>> Also, this table exists for narrowband, but still it does not for
>> wideband or ultrawideband: it would be nice to get also those ones. I
>> was really lost implementing this in my SIP application.
>
> Yes, I just checked that in into svn. Will be part of the 1.2beta2
> manual (expected soon).

And will you add thoses tables in the draft?

>>
>>    Examples:
>>
>>       m=audio 8008 RTP/AVP 97
>>       a=rtpmap:97 speex/8000
>>       a=fmtp:97 mode=4
>>
>>    This examples illustrate an offerer that wishes to receive a Speex
>>    stream at 8000Hz, but only using speex mode 4.
>>
>> Is it a recommandation or a MUST: for me, and to allow better
>> interoperability, an application is sending "mode=4" because it
>> wishes to receive "mode=4": but, in case, the remote application
>> can only send "mode=3", the receiver MUST be prepared to receive
>> ANY mode. We can't get interoperability without this and I would
>> recommand to specify that such use-case will often happen in real
>> world and that it MUST be supported.
>
> Well, the idea is what happens if all modes can't be supported for some
> reason. This is why we were saying 8 kbps (mode 3) SHOULD be supported.
> In practice, we can also strongly recommend supporting all modes, but
> I'm not sure I want to say MUST for that.

I guess you already have my idea about this: all modes should be supported
unless you know you won't have issue.

On good thing with g729 and its extension (g729 annexe b?) is that you can 
still
receive g729b if you support only g729: this is transparent (as far as I
understood it).

For speex, the modes are not transparent and thus, If I was the one
to choose, I would add in the draft: ALL MODES MUST BE SUPPORTED ON
THE RECEIVER SIDE. That's experience of real world.

The other way would be to make it transparent like g279.

>>    Several Speex specific parameters can be given in a single a=fmtp
>>    line provided that they are separated by a semi-colon:
>>
>>       a=fmtp:97 mode=any;mode=1
>>
>> No error here: just curious why you want to allow this? Wouldn't it
>> be nice to specify that the order of mode parameter is significant?
>> I guess this is what you want? (in that case, "mode=1,mode=any" might
>> be more meaningfull?)
>
> No longer sure why we had that... Albert? Greg?
>
>> More generally, I would really like to have a line specifying that
>> whatever you proposed (ptime, mode, vbr, cng), the sender could
>> use different encoder configuration for any reason (bandwidth reason
>> or lazy developper): a speex decoder don't have to be configured before
>> decoding so an application MUST be able to decode any speex stream
>> it receive provided that the sample rate was correctly negotiated.
>
> Actually, even with an incorrect sampling rate (narrowband vs wideband),
> the Speex decoder will be able to cope. Again, I totally agree with the
> idea of getting clients to accept pretty much anything,

Good.

> I'm just trying to allow that while still taking into account the fact 
> that some clients just don't have enough bandwidth or even enough 
> RAM/ROM/MIPS to handle really handle anything that is sent to them. I'm 
> definitely interested in any suggestion that can make both possible 
> though.

Make "mode" transparent! or forget about this. My own opinion...

>> today, many speex application I've seen are broken on the receiver side,
>> because they configure decoders using SDP negotiation "wish" or "static
>> configuration": providing information about this can be valuable.
>
> Not sure I understand what you mean here. Again suggestions welcome.

I mean always the same: be prepared to decode all modes, no matter what
you sent in the SDP as preference.

For example, xlite used to have a speex "quality" parameter and no
negotiation was done: if you were sending data with another mode,
the audio was not decocded. This was exactly the same issue than
the one described above for iLBC decoder in asterisk.

Aymeric

> 	Jean-Marc