- From: John Bowler <john.cunningham.bowler@gmail.com>
- Date: Thu, 27 Feb 2025 16:51:26 -0800
- To: public-png@w3.org
- Message-ID: <CAP7U39_d6Wn46dGJ2EEmnRwhFferMykBnWsk0TwGQhQT08EEXA@mail.gmail.com>
v4: cHRM chunk fix. Entirely backward compatible; existing decoders (ones which actually check, which libpng no longer does) will simply reject the chunk (harmless since the colourspace can't be represented anyway if the spec is adhered to rigorousely), new decoders will handle it. Restart markers; argued about over and over again. Fully supported at present but you can't do a complete parallel decode without knowing where the rows start and end. That needs some information which can be cracked; i.e. a cracker can use it to defeat a decoder which uses the restart marker. The crack is basically undetectable until all the preceding blocks have been decoded. If you just want to do LZ77 "inflate" in parallel no biggy; doesn't require anything new. Complete parallel decode would be nice but it's so quaint and it requires a critical chunk. Much much better to encode in blocks, not rows; much-much-much faster even though the underlying data is the same. Standard wisdom since the late '90s. John Bowler
Received on Friday, 28 February 2025 00:51:42 UTC