Having used an IBM AS/400 I was mightily annoyed when Microsoft dropped development of its database WinFS file system. Since then I'd hoped that Microsoft might have reflected on its WinFS failure and had been adventurous enough to take a minor step and at least extend the file attributing system past the paltry number of file attributes that have been in use for time immemorial but no such luck.
Extending file attributes is certainly not a proper substitute for a database file system but would have gone a long way towards making the management of files easier.
What I had in mind was to add metadata work areas to a file past its usual EOF. As these areas are extensions to a file they would follow the file from machine to machine.
One area would be used by the O/S for say version management, recording history and environment—list of users, user history, machines the file has been loaded on, additional info on file type (a more detailed version of the header bytes in a file), extended description or filename, etc., etc. The second area would be a user scratchpad area with multiple uses. Some or all of this information could be changed or reset depending on permissions/privileges.
As far as possible, files would be compatible with other OSes, these OSes could also be equipped with file managers capable of reading the additional metadata which would further aid compatibility.
Sadly, none of this has come to pass. So hopefully this new push for a database file system will succeed as it's surely needed.
That it's built on FoundationDB is a good sign I think.
Even though the article says it may be replaced at a later date, it's reasonable to expect that it's featureset/performance won't be reduced if they do make that change.
Extending file attributes is certainly not a proper substitute for a database file system but would have gone a long way towards making the management of files easier.
What I had in mind was to add metadata work areas to a file past its usual EOF. As these areas are extensions to a file they would follow the file from machine to machine.
One area would be used by the O/S for say version management, recording history and environment—list of users, user history, machines the file has been loaded on, additional info on file type (a more detailed version of the header bytes in a file), extended description or filename, etc., etc. The second area would be a user scratchpad area with multiple uses. Some or all of this information could be changed or reset depending on permissions/privileges.
As far as possible, files would be compatible with other OSes, these OSes could also be equipped with file managers capable of reading the additional metadata which would further aid compatibility.
Sadly, none of this has come to pass. So hopefully this new push for a database file system will succeed as it's surely needed.