[linux-dvb] Re: Channel database format for dvbsak

[Holger Waechtler]

Juha Yrj�l� wrote:

Hi all,

I'm starting to write a channel list for dvbsak. I can choose at
least three different formats of storing the list:

a) XML, using libxml
	+ Simple to add additional fields (e.g. VES1820 register values)
	+ Lots of parsers in different programming languages can be used
	+ Standardized
	- Depends on libxml
	- A bit hard to read

XML is not less proprietary than everything else, you always define the fields on your own - at the end no program will be able to read your channel list that does not has explicit support for your format.

If you use a tag-based format it's easy to add new fields as well, even if I don't understand what you want to do with register values in channel lists...

b) scconf (from OpenSC <http://www.opensc.org>)
	+ Simple to add additional fields
	+ Minimal effort on my part =)
	+ Readable by humans (or at least C programmers =))
	- Depends on scconf
	+ ... which isn't too large and can easily be embedded
	  in dvbsak package
	- scconf depends on flex (until someone writes
	  a lightweight replacement parser)

I don't know opensc, I'll first have to take a look on this before I can say something useful...

(Flex is no dependency, flex is a tool;)

c) Proprietary format
	- Somewhat difficult to add additional fields
	- Needs a lot of coding!
	+ No dependencies

I suppose an optimal self-defined format is much easier to code than the solutions above. I'd propose a tag-based binary format in addition to format (b).

So a) could look like this (adapted partially from dvbtune):

<transport_stream id="0x0003" orig_nid="0x1234">
  <sat_tuning_info>
    <frequency>12345</frequency>
    <symbol_rate>12345</symbol_rate>
    <polarity>h</polarity>
  </sat_tuning_info>
  <cable_tuning_info>
    <frequency>250000</frequency>
    <symbol_rate>7146</symbol_rate>
    <modulation>64-QAM</modulation>
    <ves1820_regs reg5="123" reg7="231" />
  </cable_tuning_info>
  <ter_tuning_info>
    ...
  </ter_tuning_info>
  <descriptor tag="0xab" data="123456" />

  <service id="0x0135" ca="1">
    <description type="1" provider_name="Telenor" service_name="Canal+ BL�" />
    <stream type="3" pid="0x02AC">
      <language code="fin" type="0" />
      ...
    </stream>
    <stream type="2" pid="0x0206">
    ...
    </stream>
  </service>
  <service id="0x0134" ca="1">
    ...
  </service>
</transport_stream>

I'm not too fluent in XML, so forgive me if there's something really silly. =)

(XML looks always silly to me;)

And b) would look like this:

transport_stream 0x0003 {
	original_network_id = 0x1234;
	descriptor = "ab:123456";
	descriptor = "cd:654321";
	sat_tuning_info {
		frequency = 22345;
		symbol_rate = 12345;
		polarity = h;
	}
	cable_tuning_info {
		frequency = 250000;
		symbol_rate = 7146;
		modulation = 64-QAM;
	}
	service 0x0135 {
		ca = true;
		type = 1;
		name = "Canal+ BL�";
		provider_name = "Telenor";
		stream 0x02AC {
			type = 3;
			language = "fin";
		}
	}
	...
}

As for c), any suggestions? =)

what about binary dumped and endian-corrected structs with a tag/length pair before? This is both extensible and very simple to code.

On the performance side you won't ever find a faster parser. Neither one with a smaller code size.

Maybe you don't even need endian correction if you use the format not for channel list distribution but only to save channel lists between program starts...

What do you guys think? Which should I choose? Any other formats I
should consider?

Option 2 for debugging and option 3 for real-world applications on settop boxes with limited memory.

Holger



--
Info:
To unsubscribe send a mail to [EMAIL PROTECTED] with "unsubscribe linux-dvb" as subject.