Which “application-specific purposes”? I think that is an important thing to first ask before an answer can be given.
1 Like
I don’t think BFS was ever meant to be a relational database - it set out to be a journaled filesystem with extended attributes, optional indexes on those attributes, and a query interface that searches those indexes and returns matching files.
The idea is to enable things like:
- find files where attribute X matches Y
- track results as the filesystem changes (live queries)
- present matches as if they were a folder/search result
Be did experiment with a separate database earlier on, but that was to provide the same metadata indexing/query features that BFS integrates into the filesystem. With BFS you can keep a normal hierarchical directory tree and a metadata query view over the same objects without syncing two stores, which was the issue with a separate database.
And yeah - trying to replace a database with a filesystem, or vice-versa, is mixing two different tools for different jobs.
That said, it works pretty well as a flat database for some applications.
2 Likes
Well, I do not doubt SSDs are going to be getting to be more reliable as time goes by, but I have doubt - at least for next 5-6 years - they would be getting cheaper …
Unfortunately ‘AI movement’ and their requirement for huge datacenters motivated the memory and SSD manufactuers to refactor their manufacturing capacities for their high demand … due to capitalism – so they started focusing to serve first their need - of course as their higher profit marge on such sellings. This way consumer DDR4 and DDR5 DRAMs and SSDs would be produced in reduced quantity, that concludes in higher prices for still available consumer HW in flash based devices.
I ment 5-6 years, as if these manufacturers would expand their manufacturing capabilities - they need years to build new factories – if they would want it at all. And if they woulkd do it, they wouldn’t do it for our needs, but the expanding AI hunger or AI bubble (the difference between those will turn out later).
This way the few sources of such device manufacturers, we common consumers share our interest to obtain those HW and our smaller wallet, with such giants of industrial giga corporations and service operators ( think IT and automotive and social media first and all other players who will want have a global AI tool for their service or goal) … I’m skeptic about cheaper SSDs.
Of course there’s a possible dark horse: manufacturers of Chine might discover this gap and may want to break in to this market.
Ironically, AI may solve the problems associated with capitalism and profit margins more easily than you might imagine: While corporate boardroom executives push to replace blue-collar workers with robots, the most profitable jobs to replace are white-collar! Printers are off-the-shelf components and have been mass-produced for ages, as have been powerful CPUs. The robotics in blue-collar jobs have so many more joints and pivots to control than a printer do, that the complexity of the software of a white-collar executive’s job will be more cost-effective to replace than manual labor.
As a side-note: Jesus of Nazareth’s famous quote: “In the last days, many who were last shall be first and many first, last,” was spoken in the context of a “rich, young ruler” whose greed was narrowly stronger than his desire to be righteous. Today may be such a day.