AMB – Ancient Machine Book Format
ambook.sourceforge.net
ambook.sourceforge.net
Exactly what we needed, another format!
I agree with the general sentiment, that long-term documents should be human readable. But essentially what is being designed here is just a text document with very specific formatting, that isn't even really designed to be parsed that well.
What sucks about the format proposed is that it's not going to be so easily parsed by a human/machine. I would instantly do away with the concept on index data, the hashing technique is really not going to provide any real advantage. I think this is the sort of thing that should be left to the implementer to decide when parsing the documents, rather than per-specified.
Honestly, in terms of portability, I would prefer to pick a subset of markdown. Perhaps you could apply the following restrictions:
1. Headers to use '#' and not underline feature.
2. Stick to a line wrap of 80 characters (or less if older machines require it).
3. Code blocks should use fours space indentation and not back-ticks ('```').
4. Table should be correctly white-spaced to make the formatting nice and human viewable.
5. Do away with embedded images and instead use links.
6. 7-bit readable ASCII only, no tabs (as they suggest). A tab is different on each machine (and it bugs the living hell out of me in codebases too).
7. A twist on the idea specified, "links" (to files/documents) should be unique for the first 12 characters (or an appropriate specification). This would allow them to be longer and more verbose, but also backwards compatible.
One thing I think markdown compromises on and maybe it shouldn't have is links. I think the format:
[link_text](link_location)
Should have been something like: [link_location]
Where the location is simply displayed. I feel like the idea of hiding the content of a link is deceptive - it's name should really reflect what is actually is and by displaying the link itself you encourage it to be as short as possible.Anyway, that's my 2 cents.
Markdown has a similar syntax using angle brackets, i.e., <link_location>. See https://daringfireball.net/projects/markdown/syntax#autolink.
(e.g. a friend of mine wrote a "filesystem" in BASIC on a TRS-80 Color Computer that stored 90kb on a floppy side (you flip) and also built a "robot" that could change floppies. The IBM PC was double density and heads on both sides so it stored 4 times as much, 360kb on a floppy, the early C64 split the difference around 150kb,...)
I think the hash index is cute and probably scales up to the 1.2MB floppy range that came with the PC AT.
Backticks are such an annoyance for me that when I started learning programming I dreaded having to type them and actively avoided languages that relied too much on them.
* 7 per each 3 consecutive backticks if you don't accidentally release the alt gr
[1] https://spaghettidba.com/2011/09/19/typing-the-backtick-key-...
A Microsoft Word DOCX document is really a ZIP file and if you unzip it you will find XML files, PNG files, all sorts of things you can read with ordinary tools.
Similarly a PDF file is a graph of more-or-less JSON objects which can be nested inside "filters" which support compression, encryption, etc.
The organization of numerous small subfiles makes it easy to pick one piece out to work with in RAM which would be important on a small target.
"Containers of containers (... of containers)" is a general architecture for file formats.
The 64k limit makes it a lot easier to implement an an 8-bit computer in that you can use 16-bit pointers into one of the documents, going to 24-bit or 32-bit pointers would be a hassle but BASIC programmers might use floats which would work OK until they don't.
> %b boring text follows (usually displayed grey on grey)
All in all it seems needless.