is cache-only-prefixes an nnpfs limitation?

[Tomas Olsson]

Pavel Cahyna <pcah8322 at artax.karlin.mff.cuni.cz> writes:

> sorry for replying to this old thread...
>
I don't mind, it's one of my favourites...

> I have an idea, could the 1:1 mapping stay, with the cache file becoming a
> "sparse file"?  The "holes" would fill the part of the cache file that
> does not contain valid data.
>
Well, technically yes.  And I would love it.  The difficult part is the
requiement that we should be able to read/write files larger than the
cache.  That means cache eviction of parts of files, and making new holes
in files doesn't seem to be widely supported yet.  Convince me :)

/t