[xiph-rtp] draft issues

[Ralph Giles]

A couple of minor points that need deciding:

What should a receiver do if the packet type field is set to (3)
Reserved? I vote for "MUST ignore the packet."

We have Packet type settings for packed headers (1) and the comment
header (2) but don't say anything about sending the setup or ident
headers. Is that legal? If not, what should the receiver do if it
receives them? Presumedly they could go under "raw data packet" (0).

Likewise with the optional additional headers allowed by the theora
spec.

For theora, the SDP width and height parameters define the encoded frame
size, not the viewable frame size. We should probably clarify that this
describes the decoder requirements, but that the ident header may
specify a smaller, non-multiple of 16 window for actual display.

For theora, the draft uses the frame index as the timestamp clock; the
AVT chairs recommended we follow common practice and use the 90 kHz
presentation time clock. Mike and Derf agreed here, so Luca, please make
this change.

The Configuration SDP parameter is defined as a base16-encoded version
of the packed headers. This is fine for some applications, but I remain
worried about the size issue. How about allowing this to be just the
setup header as well, as Aaron suggested. One can tell just from the
length which it is.

-r

[Luca Barbato]

Post by Ralph Giles
What should a receiver do if the packet type field is set to (3)
Reserved? I vote for "MUST ignore the packet."

I agree

Post by Ralph Giles
We have Packet type settings for packed headers (1) and the comment
header (2) but don't say anything about sending the setup or ident
headers. Is that legal? If not, what should the receiver do if it
receives them? Presumedly they could go under "raw data packet" (0).

The server has no way to send them. Sending them as raw data (type 0) is
wrong.

Post by Ralph Giles
Likewise with the optional additional headers allowed by the theora
spec.

I'd rather put them in the Packed Configuration

Post by Ralph Giles
For theora, the SDP width and height parameters define the encoded frame
size, not the viewable frame size. We should probably clarify that this
describes the decoder requirements, but that the ident header may
specify a smaller, non-multiple of 16 window for actual display.

Added a reference in the Introduction paragraph, probably it could be
worded better.

Post by Ralph Giles
For theora, the draft uses the frame index as the timestamp clock; the
AVT chairs recommended we follow common practice and use the 90 kHz
presentation time clock. Mike and Derf agreed here, so Luca, please make
this change.

Done

Post by Ralph Giles
The Configuration SDP parameter is defined as a base16-encoded version
of the packed headers. This is fine for some applications, but I remain
worried about the size issue. How about allowing this to be just the
setup header as well, as Aaron suggested. One can tell just from the
length which it is.

The SDP has to be delivered by reliable way and doesn't have size limit
from what AVT said last time was asked. Inline delivery method is
optional already, so applications may not use it if isn't possible.
Having just the Identification packet won't let you decode the packet,
so I don't see the point of allowing that.

lu