Re: APNG mailing list

On Sun, 29 Aug 2004 19:52:23 -0400, Glenn Randers-Pehrson
<[email protected]> wrote:
> One issue which seems to be "case closed" is that the result will be
> called PNG.  Since Clause 5.2 of the ISO PNG spec says that the
> signature "indicates" that the file contains a single image is
> vague enough that they can turn it into a multiple-image format.

You'll note that it wasn't me who started nitpicking about "indicates"
vs anything else.  :)  However, I think the issue here is whether the
contents of extension chunks can violate the spec, or whether they are
to be treated as blobs if the chunk type is not defined as part of the
PNG spec.  My understanding is the latter.  So, from the point of view
of the PNG spec, there is still only a single image.

> The only answer I would have is that if APNG is widely deployed then
> that statement would no longer be true.

I guess from one point of view, this is true: a tool looking at a PNG
header would have no knowledge of whether it has one frame or whether
it's an APNG with more than one frame.  However, if such a tool was
not aware of APNG, that's exactly what the file would be -- the tool
would probably munge the extra chunks and cause a mess of the
animation, but that's to be expected.  The apng/afra chunks can be
made nocopy to avoid that.  An APNG-aware app would have to do a bit
more parsing to decide whether it could or couldn't handle the file,
if it had to differentiate between animated and non-animated modes. 
Or, it could choose to just use the first frame if animation wasn't
desired.

This type of fallback is, I believe, crucial for any adoption of a
simple animated png format; it's why I'm against using a MNG header. 
It's a far easier sell to say "it's a simple extension to PNG" rather
than "it's a whole new file format" to any potential implementors.

    - Vlad

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