> Another version would be to simply keep a large dummy file and shrink/grow
> it to keep space reserved. I don't know how fast/polite/feasable that is.

I think that this would be the simplest way to do it. It wouldn't be too
bad if you had a bunch of dummy files, all the same size. Then you could
delete or add blocks as needed. Shrinking the dummy space would be almost
no cost. Growing it would be obnoxious, but you don't really need to do
that. The datastore gets full and then new items replace old items. The
datastore never really gets emptied. I suppose there is the possible case
of a huge file taking over the entire datastore and then being knocked out
by a few tiny files, leaving a big hole. But the node doesn't have to be
all that diligent about reserving space. It can periodically check for and
fill holes.

So this solution isn't sexy. Making a psuedo-filesystem that uses a single
file would be much sexier. But the only real reason for this feature is to
make sure that the node always has enough disk space. It's not a critical
feature so much as a somewhat useful feature. So I think just using dummy
files should be fine.


_______________________________________________
Freenet-dev mailing list
Freenet-dev at lists.sourceforge.net
http://lists.sourceforge.net/mailman/listinfo/freenet-dev

Reply via email to