4 ms·
There is a fundamental difference in that you can drop in a new datatype without any program like ffmpeg or gstreamer having to learn new formats. More importa
by vidarh 5y ago
There is a fundamental difference in that you can drop in a new datatype without any program like ffmpeg or gstreamer having to learn new formats.
More importantly, how many applications actually use these as their primary means of accessing file types?
Chrome doesn't learn a new image type if I drop in a file. LibreOffice doesn't. Certainly not from the same one. Gimp doesn't. There's no reason why all of those should not be able to load a new image type without updating the applications themselves, just by dropping in a datatype/codec.
The concept isn't anything magical - or new -, but it's still largely forgotten and ignored outside of small islands of use.
We have all kinds of little API's to do this for specific types of files, like e.g. ImageMagick's "convert" for images, ffmpeg for video etc. as you mention, but what we lack is a unified API, the ability to drop plugins in for new file types, and for applications to actually unify around using a specific standard.
Heck, we even have a workaround to avoid needing application support: A FUSE filesystem to trigger conversion to suitable target filetypes.
This is the same issue with a lot of the things we're missing from AmigaOS. They're not hard to do. E.g. "assigns" are conceptually largely a OS-provided way of creating PATH's for everything you'd like, for example - it can also easily be provided via a FUSE filesystem. Dbus "almost" gets us to what AREXX provided, but is just not supported as widely or as simple for regular users to get started with.
What was special with the Amiga in that respect was that these things were embraced fully across the board, and integrated into the culture surrounding the Amiga.
- nitrogen 5y agounified API, the ability to drop plugins in for new file types, and for applications to actually unify around using a specific standard. There have been attempts, such as OLE and ActiveX on Windows, or KParts in KDE 2 (and dcop from Qt 2) (or was it 3?), or Intents on Android, or WebIntents, but yeah, none of them became universal, and many of them had security problems.
- vidarh 5y agoI think all of these illustrate part of the challenge: They're all tremendously complex compared to the basic functionality of datatypes. That is part of the problem with many of the attempts to do the same again. A lot of the constraints AmigaOS had to deal with became strengths in that these APIs had to be simple because they had to work on machines with memory measured in kilobytes or megabytes. Of course it's also more complex to do on a system where you expect security (no memory protection on Amiga's made that rather moot), but I think a targeted attempt to do just the minimum datatypes does - or even less, would still be valuable. E.g. many of the security issues are related to embedding a widget, but just providing a unified mechanism for loading and saving which could easily be sandboxed and interfaced to via a dumb pipe (still security issues in validating what's passed over that of course).
- AnIdiotOnTheNet 5y agoIndeed. Systems like the Amiga show us what we have lost from the world of personal computing. Things we could do in the 80s and 90s have disappeared and been replaced by hodgepodges of barely interacting components. It is, frankly, disgraceful and we should be ashamed as an industry that we've regressed so far.
- pjmlp 5y agoThe idea lives on as COM, XPC and DBUS. The problem is that many kind of ignore using them.
- tialaramex 5y ago> There is a fundamental difference in that you can drop in a new datatype without any program like ffmpeg or gstreamer having to learn new formats. It's not an interesting or useful difference. The thing Datatypes lets you do is pointlessly spell things differently. Once you spot that you ask yourself, "Why am I doing this?" and the answer is only that you need to interoperate with other people who hadn't settled on one way to spell things so the best route forward is consolidation. Most applications gain nothing from having, say, yet another mediocre bitmap graphics format. You just made your system more complicated without any improvements whatsoever. Clearly we should agree to just stop doing this. And we did. > Gimp doesn't Actually Gimp does. Gimp's file format handling is all via plug-ins. As with ImageMagick in this context maybe you actually needed to rescue crappy 1980s files you found on a floppy disk or whatever. But this helps us see what's going on - because Gimp's image file format handling is only for import and export of necessity Gimp itself is richer than any of these formats. Its native XCF format reflects a superset of the features you can care about. But anything Gimp's native images can't do, you can't "add" using these plug-ins. The SVG plug-in for example can't magically make Gimp understand vector images, it just has to render the vector image as pixels because that's all Gimp understands. So in the end any innovation is rendered pointless by this abstraction, the only way for an innovative idea to flourish is to sidestep a "datatypes"-like abstraction altogether. > actually unify around using a specific standard. You've almost understood. Instead of adding a layer of abstraction, and then unifying around this unnecessary layer, we instead chose to just use a specific standard format. The problem you're intent on solving vanished. Didn't you ever wonder why Datatypes seemed so brilliant for image files, and kinda sorta OK for music, and then basically of no value at all for most other file formats? It's because there really were sixty different crappy pixmap formats that were completely interchangeable and half a dozen PCM audio formats likewise. But that's not what happened in other domains, because it's pointless.
- vidarh 5y ago> Most applications gain nothing from having, say, yet another mediocre bitmap graphics format. Applications gain a tremendous amount of usability for me by being able to read or write the formats I actually use. Instead we rely on a hodgepodge of scripts and tools to do format conversion manually. Amiga applications released decades ago can load webp files for example, whether or not the people who wrote them are still releasing updates or are even around. It might not matter to you. It matters to me that files keep being easily accessible. Today there are plenty of tools I use regularly that can't load or save various formats I regularly use directly. That is pointless, when providing a mechanism to reuse conversion was a solved problem decades ago. > Actually Gimp does. Gimp's file format handling is all via plug-ins. But it's not reusing system-wide plugins, which was the point. > You've almost understood. Instead of adding a layer of abstraction, and then unifying around this unnecessary layer, we instead chose to just use a specific standard format. The problem you're intent on solving vanished. Unfortunately you failed to understand. We haven't unified on a standard format. We never will, because we get new requirements regularly, and people also create new formats for arbitrary and stupid reasons as well. Format conversion will always be necessary to handle formats produced by different tools. We can choose to do that "out of band", or we can choose to re-implement it for applications time and time again, or we can move towards a unified system. Today we do the first two. A tremendous amount of time is wasted implementing conversion and loading and saving for different applications, and a tremendous amount of time is wasted on doing out-of-band conversion. I've spent the better part of the last month dealing with crappy data conversion because a tool I need to feed data into implements it's own conversion in a badly broken way instead of reusing a better one, forcing me to go via another format. The amount of developer effort wasted on this bullshit is massive, and the amount of inconvenience for users is too. > Didn't you ever wonder why Datatypes seemed so brilliant for image files, and kinda sorta OK for music, and then basically of no value at all for most other file formats? It's because there really were sixty different crappy pixmap formats that were completely interchangeable and half a dozen PCM audio formats likewise. But that's not what happened in other domains, because it's pointless. No, I didn't wonder that, because the limitation there was that Commodore went bankrupt and datatypes was not progressed to cover additional domains, and the Amiga community was not large enough to take on the task of covering more complex structured data options. Meanwhile we still depend on conversion filters for all kinds of other formats all the time for e.g. office suites. So no, it's not pointless.
- rbanffy 5y ago> What was special with the Amiga in that respect was that these things were embraced fully across the board That’s just because it’s part of the OS. On a Mac you have lots of high level APIs to interact with programs and data that form the visible layers of the OS. On Linux there is a hard border between GUI and OS and a mechanism like this would be more at home in Gnome or other desktops. Another factor that drove adoption is software scarcity. Back then it wasn’t trivial to just link to a library that reads GIF files and it was much easier to use OS services for that. The cost comes in portability. I can write a Mac or Linux app that can be compiled and runs on mostly anything, but the moment I use OS services, it’s no longer portable.
- vidarh 5y agoThe thing is there's nothing preventing making a portable API providing the most important capabilities datatypes did. It's one tiny little library. The challenge is cultural, not technical. Without a reasonable hope of adoption, there's little value. It's also not as simple as it being part of the OS, though space constraints certainly would favour OS libraries - the Amiga world is full of APIs that were adopted across the board despite not being part of the OS. E.g. XPK is a good example: If you want to support compression in an Amiga app, you'll most likely use XPK, which gives you instant support for every compression algorithm someone wrote an XPK library for. What is more likely to have mattered was in general whether or not a potential user would be likely to already have a library installed. My first hard drive was 20MB. What an application chose to install was a big deal, and that certainly encouraged people think about sticking to options that were popular and that likely to would save space. Regarding the "hard border", it's worth keeping in mind that while datatypes could be used for visual elements, it could also be used for "headless" load/save/conversion. Putting something like it in e.g. Gnome would be less valuable than providing the "headless" part, because the format conversion is the most valuable aspect. And that part is incidentally also the easiest part to make portable.
- rbanffy 5y agoI mentioned Gnome because it’s an application platform as much as it’s a graphical environment. You write apps for it using its services to integrate with the desktop. Datatypes don’t make as much sense at the lower OS level because the OS is not supposed to be aware of the data types applications work with.
- pjmlp 5y agoWindows has followed up the idea with COM, for example, the datatypes version would be WIC. https://docs.microsoft.com/en-us/windows/win32/wic/-wic-lh https://docs.microsoft.com/en-us/windows/win32/wic/-wic-lh However it is exactly as you put it, it is exposed at application level, so unless an application calls into WIC, it will never be made aware of new WIC components deployed into the system.