3 ms·

This post is full of interesting new ideas. I'd love to see these implemented in a mainstream Linux distro (or Windows/Mac), but the backward compatibility cons

This post is full of interesting new ideas. I'd love to see these implemented in a mainstream Linux distro (or Windows/Mac), but the backward compatibility considerations are a huge impediment. Literally all the software in the past 4 decades (at least) has used directory and file based file systems. One way to encourage usage and let developers find advantages of the new system could be to offer this as an alternative file system that can be used in parallel with existing ones. That would also uncover bugs/problems in the new system which could be fixed. PS: I do think the name "didget" could be improved. It is not as natural to write or speak as "file".

theamk 5y ago
Is it really full of new ideas? This looks like a fairly generic tagging system which seems to pop up on HN every few months. If you are interesting in tagging systems, I recommend you to look at Microsoft WinFS (part of Longhorn). This took idea of tags very deeply -- not only it had user-specified tags, but it would parse many existing file formats and automatically generate tags based on contents -- for example, if it saw a word doc and detected it is of "resume" type, it'd extract "Name, Educational Qualification, Experience" values from it. Sadly, there is no good single overview I know of. You can start from wikipedia (https://en.wikipedia.org/wiki/WinFS https://en.wikipedia.org/wiki/WinFS) and follow the links.
didgetmaster 5y ago
I use the name Didget (short for Data Widget) because it encompasses much more than file data. A 'File Didget' is an object where its data stream has unstructured data and is unmanaged by the system. But there are other Didgets (Schema Didgets, Set Didgets, Tag Didgets, etc.) where their data stream contents are completely managed. Like NTFS that uses files for its internal data structures and many databases that use relational tables to manage databases, I chose to enclose all my internal structures (e.g. tags, file tables, allocation bitmaps, etc.) within other Didgets.

theamk 5y ago

Is it really full of new ideas? This looks like a fairly generic tagging system which seems to pop up on HN every few months. If you are interesting in tagging systems, I recommend you to look at Microsoft WinFS (part of Longhorn). This took idea of tags very deeply -- not only it had user-specified tags, but it would parse many existing file formats and automatically generate tags based on contents -- for example, if it saw a word doc and detected it is of "resume" type, it'd extract "Name, Educational Qualification, Experience" values from it. Sadly, there is no good single overview I know of. You can start from wikipedia (https://en.wikipedia.org/wiki/WinFS https://en.wikipedia.org/wiki/WinFS) and follow the links.

didgetmaster 5y ago

I use the name Didget (short for Data Widget) because it encompasses much more than file data. A 'File Didget' is an object where its data stream has unstructured data and is unmanaged by the system. But there are other Didgets (Schema Didgets, Set Didgets, Tag Didgets, etc.) where their data stream contents are completely managed. Like NTFS that uses files for its internal data structures and many databases that use relational tables to manage databases, I chose to enclose all my internal structures (e.g. tags, file tables, allocation bitmaps, etc.) within other Didgets.