Is using BFS file attributes for units really worth it?

[Teknomancer]

Originally it was a cool idea to use BFS attributes to store the unit conversions.

It seemed the BeOS way of doing things. However, in reality does it bring any practical benefits?

Arguments for using a regular file (like TOML):

1. Every unit will not have to be stored as a separate file. Otherwise, it might involve a ton of 'stat' or similar syscalls that actually slows things down versus opening just one or two files.
2. Storing the units in the source code repository is a bit of hassle because they will need to be compressed and stored as binaries. Every change required will require re-uploading entire binaries (zip file/s) and changes made often will quickly multiply the size of the project.
3. Adding or modifying units to the base application (i.e. the source tree) will require a Haiku system or VM.

Arguments for using BFS attributes:

1. Units are universal standards and don't change that often.
2. A cool way of show casing the BFS capability.
3. File size of the zip of all units won't be that large and isn't a great cause for concern in the foreseeable future.

Of course, changing to using a regular file for this requires quite a bit of code change but in the long run I feel it might be the better solution.

Regardless, a lof of changes will be required anyway given we are reviving a dusty relic that has not been maintained for 20 years.

[PeteCA]

Personally I always thought that using Attributes was a really neat scheme! The fact that it would have to be a zip file here is a bit of a pain, but not a biggie. As units essentially don't change it only has to be created once. Adding more units can just be a separate zip. I never noticed a speed problem even on old slower BeOS. As it will only run on Haiku, that environment would seem to be a requirement anyway.