This worries me. Because presumably, changing the compression algorithm will break backwards compatibility, which means we'll start to see "png" files that aren't actually png files.
It'll be like USB-C but for images.
That being said, they also can do dumb things however, right at the end of the sentence you quote they say:
> we want to make sure we do it right.
So there's hope.
[1] https://github.com/w3c/png/issues/39#issuecomment-2674690324
That's just changing an implementation detail of the encoder, and you don't need spec changes for that e.g. there are PNG compressors which support zopfli for extra gains on the DEFLATE (at a non-insignificant cost). This is transparent to the client as the output is still just a DEFLATE stream.
The PNG format is specifically designed to allow software to read the parts they can understand and to leave the parts they cannot. Having an extensible format and electing never to extend it seems pointless.
This proves OP analogy regarding USB-C. Having PNG as some generic container for lossless bitmap compression means fragmentation in libraries, hardware support, etc. The reason being that if the container starts to support too many formats, implementations will start restricting to only the subsets the implementers care about.
For instance, almost nobody fully implements MPEG-4 Part 3; the standard includes dozens of distinct codecs. Most software only targets a few profiles of AAC (specifically, the LC and HE profiles), and MPEG-1 Layer 3 audio. Next to no software bothers with e.g. ALS, TwinVQ, or anything else in the specification. Even libavcodec, if I recall correctly, does not implement encoders for MPEG-4 Part 3 formats like TwinVQ. GP's fear is exactly this -- that PNG ends up as a standard too large to fully implement and people have to manually check which subsets are implemented (or used at all).
> Many of the programs you use already support the new PNG spec: Chrome, Safari, Firefox, iOS/macOS, Photoshop, DaVinci Resolve, Avid Media Composer...
It might be too late to rename png to .png4 or something. It sounds like we're using the new png standard already in a lot of our software.
EG your GPU and monitor both have a USB-C port. Plug them together with the right USB cable and you'll get images displayed. Plug them together with the wrong USB cable and you won't.
USB 3 didn't have this issue - every cable worked with every port.
If you've created an extensible file format, but you never need to extend it, you've done everything right, I'd say.
I believe the problem here is that you will have PNG images that “look” like you can open them but can’t.
What was broken was the promise of a "single cable to rule them all", partly due to manufacturers ignoring the requirements of USB-C (missing resistors or PD chips to negotiate voltages, requiring workarounds with A-to-C adapters), and a myriad of optional stuff, that might be supported or not, without a clear way to indicate it.
If PNG gets extended, it's entirely plausible that someone will view a PNG in their browser, save it, and then not be able to open the file they just saved.
There are those who claim "backwards compatibility" doesn't cover "how you use it" - but roughly none of the people who now have to deal with broken software care about such semantic arguments. It used to work, and now it doesn't.
Same is also true for the most advanced codecs. MPEG-* family and MP3 comes to my mind.
Nothing stops PNG from defining a "set of decoders", and let implementers loose on that spec to develop encoders which generate valid files. Then developers can go to town with their creativity.
Do they mention which C libraries use this spec?
That's what I would call really extensible, but then there may be no limits and hacking/viruses could have easily a field day.
In an ideal world, yes. In practice however, if some field doesn't change often, then software will start to assume that it never changes, and break when it does.
TLS has learned this the hard way when they discovered that huge numbers of existing web servers have TLS version intolerance. So now TLS 1.2 is forever enshrined in the ClientHello.
And considering we already have plenty of more advanced competing lossless formats, I really don't see why "feed a BMP to deflate" needs a new, incompatible spin in 2025.
Will sooner or later be used to implement RCEs. Even if you could do a restriction as is done for eBPF, that code still has to execute.
Best would be not to extend it.
This is just pretending that if you have a cat and a dog in two bags and you call it “a bag”, it’s one and the same thing…
Regarding the potential for fragmentation of the png ecosystem the alternative is a new file format which has all the same support issues. Every time you author something you make a choice between legacy support and using new features.
From a developer perspective, adding support for a new compression type is likely to be much easier than implementing logic for an entirely new format. It's also less surface area for bugs. In terms of libraries, support added to a dependency propagates to all consumers with zero additional effort. Meanwhile adding a new library for a new format is linear effort with respect to the number of programs.
Other than JXL which still has somewhat spotty support in older software? TIFF comes to mind but AFAIK its size tends to be worse than PNG. Edit: Oh right OpenEXR as well. How widespread is support for that in common end user image viewer software though?
USB-C spec is anything but breaking backward compatible.
Labelling is a poor band-aid on the root problem - consumer cables which look identical and fit identically should work wherever they fit.
There should never have been a power-only spec for USB-C socket dimensions.
If a cable supports both power and data, it must fit in all sockets. If a cable supports only power it must not fit into a power and data socket. If a cable supports only data, it should not fit into a power and data socket.
It is possible to have designed the sockets under these constraints, with the caveat that they only go in one way. I feel that that would have been a better trade-off. Making them reversible means that you cannot have a design which enforces cable type.
It's a dichotomy. Either the provider accommodates users with older software or not. The file extension or internal headers don't change that reality.
Another example, new versions of PDF can adopt all the bells and whistles in the world but I will still be saving anything intended to be long lived as 1/a which means I don't get to use any of those features.
Not Sure what youre talking abouz.
If you want to check yours: mediainfo **/*.mp4 | grep -A 2 '^Audio' | grep Format | sort | uniq -c
https://en.wikipedia.org/wiki/TwinVQ#TwinVQ_in_MPEG-4 tells the story of TwinVQ in MPEG-4.
The main use case for PNG is web browsers and all of them seem to be on board. Using old web browsers is a bad idea. You do get these relics showing up using some old version of internet explorer. But some images not rendering is the least of their problems. The main challenge is actually going to be updating graphics tools to export the new files. And teaching people that sRGB maybe isn't good enough any more. That's going to be hard since most people have no clue about color spaces.
Anyway, that gives everybody plenty of time to upgrade. By the time this stuff is widely used, it will be widely supported. So, you kind of get forward compatibility that way. Your browser already supports the new format. Your image editor probably doesn't.
And now think of the younger generation that has grown up with smartphones and have been trained to not even know what a file is. I remember this story about senior high school students failing their school tests during covid because the school software didn't support heif files and they were changing the file extension to jpg to attempt to convert them.
I have no trust the software ecosystem will adapt. For instance the standard libraries of the .net framework are fossilised in the world of multimedia as of 2008-ish. Don't believe heif is even supported to this day. So that's a whole bunch of code which, unless the developers create workarounds, will never support a newer png format.
https://svgees.us/blog/img/revoy-cICP-bt.2020.png uses the new colour space. If your software and monitor can handle it, you see better colour than I, otherwise, you see what I see.
That's even more confusing than the current state of affairs. If my phone has power and data socket, then I cannot use power only cable to only charge it? Presumably with the charger that has power only socket. So I need a cable with two different ends anyway. Just go micro-USB at this point :)
Funnily enough, there is a 100% overkill way to solve such issues. Just use super expensive certified TB cables. Well... plus a A-to-C adapter for noncompliant devices, I guess.
More generally, PNG has a simple feature to specify what's needed. A file consists of a number of chunks, and one bit in the chunk specifies whether that chunk is required for display. All of the extensions I've seen in the past decades set that bit to "optional".
For example, this update includes a chunk containing EXIF data. As you'd expect, the exif chunk sets that bit to "optional".
Also if you forbid evolving existing formats, the only alternative to improve is to introduce a new format, and I argue that it would be causing even more fragmentation and be more difficult to adopt to. Look at all the drama surrounding JPEG XL.
Well, yes.
Why can't you use a power+data cable for the vape (or whichever appliance takes both)? What's the deal-breaker here?
The alternative is labeling, or plugging cables in to see if they do what you want them to do.
Both are a poor user interface.
I’m not saying this is what will happen — but if I was able to construct a plausible approach to compression in ten minutes, then perhaps it’s a bit early to predict the doom of compatibility.
But that's typical for file extensions. Consider EXE – it is probably an executable, but an executable for what? Most commonly Windows – but which Windows version will this EXE run on? Maybe this EXE only works on Windows 11, and you are still running Windows 10. Or maybe you are running x86-64 Windows, but this EXE is actually for ARM or MIPS or Alpha. Or maybe it is for some other platform which uses that extension for executable files – such as DOS, OS/2, 16-bit Windows, Windows CE, OpenVMS, TOPS-10, TOPS-20, RSX-11...
.html, .js, .css – suggest to use a web browser, but don't tell you whether they'll work with any particular one. Maybe they use the latest features but you use an old web browser which doesn't support them. Maybe they require deprecated proprietary extensions and so only work on some really old browser. Maybe this HTML page only works on Internet Explorer. Maybe instead of UTF-8 it is in some obscure legacy character set which your browser doesn't support.
.zip – supports extensible compression and encryption methods, your unzip utility might not support the methods used to compress/encrypt this particular zip file. This is actually normal for very old ZIP files (from the 1980s) – early versions of PKZIP used various deprecated compression mechanisms, which few contemporary unzip utilities support. The format was extended to 64-bit without changing the extension, there's still a lot of 32-bit only implementations out there. ZIP also supports platform-specific file attributes–e.g. PKZIP for z/OS creates ZIP files which contain metadata about mainframe data storage formats, unzip on another platform is going to have no idea what it means, but the metadata is actually essential to interpreting the data correctly (e.g. if RECFM=V you need to parse the RDWs, if RECFM=F there won't be any)
.xml - okay, it is XML – but that tells you nothing about the actual schema. Maybe you were expecting this xml file to contain historical stock prices, but instead it is DocBook XML containing product documentation, and your market data viewer app chokes on it. Or maybe it really is historical stock prices, but you are using an old version of the app which doesn't support the new schema, so you can't view it. Or maybe someone generated it on a mainframe, but due to a misconfiguration the file came out in EBCDIC instead of ASCII, and your app doesn't know how to read EBCDIC, yet the mainframe version of the same app reads it fine...
.doc - people assume it is legacy (pre-XML) Microsoft Word: every version of which changed the file format, old versions can't read files created with newer versions correctly or at all, conversely recent versions have dropped support for files created in older versions, e.g. current Office versions can't read DOC files created with Word for DOS any more... but back in the 1980s a lot of people used that extension for plain text files which contained documentation. And it was also used by incompatible proprietary word processors (e.g. IBM DisplayWrite) and also desktop publishing packages (e.g. FrameMaker, Interleaf)
.xmi – I've seen this extension used for both XML Model Interchange (XML-based standard for exchanging UML diagrams) and XMIT (IBM mainframe file archive format). Because extensions aren't guaranteed to be unique, many incompatible file formats share the same extension
.com - is it an MS-DOS program, or is it DCL (Digital Command Language)?
.pic - probably some obscure image format, but there are dozens of possibilities
.img – could be either a disk image or a visual image, either way dozens of incompatible formats which use that extension
.db – nowadays most likely SQLite, but a number of completely incompatible database engines have also used this extension. And even if it is SQLite, maybe your version of SQLite is too old to read this file because it uses some features only found in newer versions. And even if SQLite can read it, maybe it has the wrong schema for your app, or maybe a newer version of the same schema which your old version that app doesn't support, or an old version of the schema which the current version of the app has dropped support for...
This is news to me. I'm pretty sure the main use case for PNG is lossless transparent graphics.
Proprietary or open, any visual codec is a battleground. Even in commercial settings, I vaguely remember people saying they prefer the end result of one encoder over another, for the same video/image format, not unlike how photographers judge cameras by their colors.
So maybe, this flexibility to PNG will enable or encourage people to write better or at least unorthodox encoders which can be decoded by standard compliant ones.
It's not, most images you encounter on the web need better compression.
The main PNG use case is to store lossless images locally as master copies that are then compressed or in workflows where you intend to edit and change them where compressed formats would degrade the more they were edited.
Under Windows 95/98/Me, most command line tools were MS-DOS executables. Their support for 32-bit Windows console apps was very poor, to the extent that the input and output of such apps was proxied through a 16-bit MS-DOS executable, conagent.exe
First time in my life I ever used GNU Emacs, it was an OS/2 exe. That's also true for bash, ls, cat, gcc, man, less, etc... EMX was my gateway drug to Slackware
There are about 3.6 billion people surfing the web and experiencing PNGs. That use case, consuming PNGs, seems to dwarf the perhaps 100 million (somewhat wild guess) graphic designers, web developers, and photo editing professionals who manipulate images for publishing (in any medium) or archiving.
If, on the other hand, you're considering the use cases envisioned by PNG's creators, or the use cases that interest the people processing or publishing images, yes, these people are focused on format itself and its capabilities.
I suspect this particular use of "use case" isn't terribly clear. Also these two considerations are not incompatible.
The first bit of our research is "What can we already make use of which requires no spec update? There are plenty of PNG optimizers. How much of that should go into the typical PNG libraries?"
Same with parallel encoding & decoding. An older image viewer will be able to decode it on one thread without ever knowing parallel decoding was an option.
Here's the worry-a-little part: Everybody immediately jumps to file size as to what image compression is better or worse. That isn't the best take, but it is what it is. So there is pressure to adopt newer technologies.
We often do have a way to maintain some degree of backwards compatibility even when we do this. For example, we can store a downsampled image for old viewers. Then extra, new chunks will know "Mix that with this full scale data, using a different compression".
As you can imagine, this mixing complicates things. It might not be the best option. Sooooo we're researching it :)
Did you know that Microsoft Windows originally ran on top of the much older MS-DOS, which used EXE files as one of its two executable formats? Most Windows users had lots and lots of EXE files which were not Windows executables, but instead DOS executables. And then came Windows 95, which introduced 32-bit Windows executables, but kept the same file extension as 16-bit Windows executables and 16-bit DOS executables.
You don't follow spec, you're on your own.
https://pico-8.fandom.com/wiki/P8PNGFileFormat
Actual cases of proprietary chunks include iDOT from Apple (apparently a performance optimization for plain images)
https://www.hackerfactor.com/blog/index.php?/archives/895-Co...
and the Macromedia Fireworks save files
https://stackoverflow.com/questions/4242402/the-fireworks-pn...
