Re: Ogg encapsulation

On Thu, Apr 01, 2004 at 09:05:54PM +0200, Gerard Juyn wrote:

> > > mng_retcode mng_set_checkcrc (mng_handle hHandle,
> > >                                    mng_uint8 iMode);
> > > iMode can be:
> > > 0x00 = crc in datastream & must be checked (default)
> > > 0x01 = crc in datastream, must not be checked
> > > 0x02 = no crc in datastream, no checking

I think you need to distinguish 'should check crc' in the we-already-
know-it's-valid sense from 'no crc in the datastream'. The latter is 
really a special feature for applications like mine where the datastream 
isn't as in the spec.

> The bits in iMode (now a uint16) become flags:
> 
> 0x0001 crc present (or not)

crc present in datastream (or not)

> 0x0010 crc must be checked for ancillary chunks (or not)

crc should be checked...

> 0x0020 faulty crc on ancillary chunks generate a warning (or not)
> 0x0040 faulty crc on ancillary chunks generate an error (or not)

There should also be a note that these fields are ignored if the crc is 
not present/not checked.

I think that makes the intent of the flags more clear.

> Actually, errors can be ignored as well by #define MNG_SOFTERRORS
> This will change the error-handling macros to work like a warning. Sure hope 
> people know what they're doing if they start messing with that, as it means 
> the default will then be to ignore errors!!!

Quite. I wouldn't worry about abuse particularly. Lazy programmers 
can find much easier ways of breaking the format. Just look at the 
history of png support in Photoshop or Internet Exporer.

 -r


--
Send the message body "help" to [email protected]