In my opinion it is all wrong. This loop deals with legality of file, not how the demuxer reads the file. Current:
file: file_id_string while(!eof){ packet_header, main_header, packet_footer reserved_headers for(i=0; i<stream_count; i++){ packet_header, stream_header, packet_footer reserved_headers } while(next_code == info_startcode){ packet_header, info_packet, packet_footer reserved_headers } if(next_code == index_startcode){ packet_header, index_packet, packet_footer } if (!eof) while(next_code != main_startcode){ if(next_code == syncpoint_startcode){ packet_header, syncpoint, packet_footer } frame reserved_headers } }
The way it should be in my opinion: file: file_id_string while(!eof){ packet_header, main_header, packet_footer for(i=0; i<stream_count; i++){ packet_header, stream_header, packet_footer } reserved_headers while(next_code == info_startcode){ packet_header, info_packet, packet_footer } reserved_headers if(next_code == index_startcode){ packet_header, index_packet, packet_footer } if (!eof) while(next_code != main_startcode){ packet_header, syncpoint, packet_footer do { frame } while (next_code == framecode) reserved_headers if (realtime_stream) { while(next_code == info_startcode){ packet_header, info_packet, packet_footer } reserved_headers } } } 1. no reserved headers allowed until end of stream headers this offers absoloutely no disadvantage in extendibility: a. new packets can still be written afterwards. b. if it is so important that it needs to be written either next to the main header or next to the streams, it can be written inside the main header or inside the stream headers (reserved_bytes) 2. no reserved headers allowed after index - only either an eof or a syncpoint and frames. no reserved headers inbetween info packets. 3. syncpoints must be written after any non frame packets, before any frame packets. 4. it is illegal to write a syncpoint without a following frame.. 5. without a main header set, only reserved headers, info packets, and syncpoints can be written (this solves the question i had in previous mail). Rational: a demuxer will give all info packets after the main header set togther with reading the main header set to the caller, ONCE. Any info packets seen after a main header set afterwards are ignored. Any info packets seen OUTSIDE of a main header set are "new information" packets and given to the caller on sight... These packets would have a manditory chapter_id!=0, and are illegal in non-realtime streams. 6. as before, main headers can be written again after main headers.. I'm sorry for bringing up this issue so late :( - ods15
Hi On Sat, Nov 04, 2006 at 04:30:49PM +0200, Oded Shimon wrote:
In my opinion it is all wrong. This loop deals with legality of file, not how the demuxer reads the file.
Current:
file: file_id_string while(!eof){ packet_header, main_header, packet_footer reserved_headers for(i=0; i<stream_count; i++){ packet_header, stream_header, packet_footer reserved_headers } while(next_code == info_startcode){ packet_header, info_packet, packet_footer reserved_headers } if(next_code == index_startcode){ packet_header, index_packet, packet_footer } if (!eof) while(next_code != main_startcode){ if(next_code == syncpoint_startcode){ packet_header, syncpoint, packet_footer } frame reserved_headers } }
The way it should be in my opinion: file: file_id_string while(!eof){ packet_header, main_header, packet_footer for(i=0; i<stream_count; i++){ packet_header, stream_header, packet_footer } reserved_headers while(next_code == info_startcode){ packet_header, info_packet, packet_footer } reserved_headers if(next_code == index_startcode){ packet_header, index_packet, packet_footer } if (!eof) while(next_code != main_startcode){ packet_header, syncpoint, packet_footer do { frame } while (next_code == framecode) reserved_headers if (realtime_stream) { while(next_code == info_startcode){ packet_header, info_packet, packet_footer } reserved_headers } } }
1. no reserved headers allowed until end of stream headers this offers absoloutely no disadvantage in extendibility: a. new packets can still be written afterwards. b. if it is so important that it needs to be written either next to the main header or next to the streams, it can be written inside the main header or inside the stream headers (reserved_bytes)
no, there are cases where it makes sense think about error correcting packets (with reed solomon parity symbols for example) yes its a silly example but its not entirly unrealistic and such error correcting packets pretty much must be after the packet they apply to as the info is needed for parsing of later packets placing the stuff in the packet might cause some problems with future extendability ... another similar example would be a signature to proof that a header hasnt been tampered with ... also what is the problem with allowing such extra headers there? just disallowing something because we dont see an immedeate use is not good design the same points apply to the other cases where you want reserved_headers removed [...]
3. syncpoints must be written after any non frame packets, before any frame packets.
i think that is required by other parts of the spec already [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB In the past you could go to a library and read, borrow or copy any book Today you'd get arrested for mere telling someone where the library is
On Sun, Nov 05, 2006 at 12:42:39AM +0100, Michael Niedermayer wrote:
Hi
On Sat, Nov 04, 2006 at 04:30:49PM +0200, Oded Shimon wrote:
The way it should be in my opinion: file: file_id_string while(!eof){ packet_header, main_header, packet_footer for(i=0; i<stream_count; i++){ packet_header, stream_header, packet_footer } reserved_headers while(next_code == info_startcode){ packet_header, info_packet, packet_footer } reserved_headers if(next_code == index_startcode){ packet_header, index_packet, packet_footer } if (!eof) while(next_code != main_startcode){ packet_header, syncpoint, packet_footer do { frame } while (next_code == framecode) reserved_headers if (realtime_stream) { while(next_code == info_startcode){ packet_header, info_packet, packet_footer } reserved_headers } } } [...] 3. syncpoints must be written after any non frame packets, before any frame packets.
i think that is required by other parts of the spec already
Yes, I know, I wanted it written in the main loop. OK, I accept your arguments on reserved data. I still have some thoughts on info packets, will reply in other mail.. - ods15
participants (2)
-
Michael Niedermayer -
Oded Shimon