probably you should create a partition with 1KB block size
2003-09-14 25746, 2003
yalaforge
otherwise you waste 3K for each file
2003-09-14 25748, 2003
djce
Six months ago or so, it was 3319 Mb
2003-09-14 25700, 2003
djce
Good idea. Do you know how you do that?
2003-09-14 25724, 2003
yalaforge
you need a free partition, one you don't need anymore
2003-09-14 25735, 2003
djce
I was thinking of seeing how well it would fit into a PG database too...
2003-09-14 25746, 2003
djce
OK. Or I could use a "-o loop" file, yes?
2003-09-14 25747, 2003
yalaforge
use mke2fs with the -b option
2003-09-14 25751, 2003
djce
Ah, ok.
2003-09-14 25708, 2003
djce
I don't do a lot of that normally. The OS installer takes care of that for me :-)
2003-09-14 25710, 2003
yalaforge
I don't know about loopback file systems, sorry
2003-09-14 25713, 2003
djce
ok
2003-09-14 25729, 2003
yalaforge
you need to umount the partition in question
2003-09-14 25743, 2003
yalaforge
but all data on it will be lost, of course.
2003-09-14 25755, 2003
djce
yes, of course. thanks.
2003-09-14 25702, 2003
yalaforge
mke2fs -b 1024 -j /dev/hd...
2003-09-14 25709, 2003
yalaforge
oh, and you need enough inodes. At least one for each file (the -N option)
2003-09-14 25719, 2003
djce
1067811
2003-09-14 25722, 2003
djce
ouch!
2003-09-14 25706, 2003
yalaforge
oh dear. that's a lot of files
2003-09-14 25744, 2003
yalaforge
with 1K block size it should be just a bit over 1Gig
2003-09-14 25710, 2003
yalaforge
why do you need all that crap?
2003-09-14 25711, 2003
djce
I'm just gonna see how big the tar file is after "bunzip2"
2003-09-14 25726, 2003
djce
I don't. It's just for, err, fun. Or interest. Or because I can.
2003-09-14 25735, 2003
yalaforge
:-)
2003-09-14 25704, 2003
djce wonders how much faster a simple decompress will be than a decompress-and-extract.
2003-09-14 25733, 2003
yalaforge
much faster
2003-09-14 25747, 2003
djce
We'll see. But yet, I think so!
2003-09-14 25751, 2003
djce
(yes)
2003-09-14 25705, 2003
yalaforge
the problem with extract is that many new data structures have to be created in the file system
2003-09-14 25730, 2003
yalaforge
ext2fs doesn't cope very well with this. The BSD FS in async mode is even worse
2003-09-14 25709, 2003
djce
Yup. I'm hoping with just a bunzip2 (i.e. freedb-complete.tar) that I can scan the tar file, build up an index of filenames to TAR offsets, and then just "open/seek/read" to load each record.
2003-09-14 25726, 2003
djce
Hopefully, seek is nice and fast.
2003-09-14 25735, 2003
yalaforge
yes, I think that'll work
2003-09-14 25701, 2003
yalaforge
I once fed a tar file full of mp3s to my xmms. It worked well :-)
2003-09-14 25747, 2003
yalaforge
cool, there is a description of the tar format in /usr/include/tar.h
2003-09-14 25713, 2003
djce
bunzip2: 10 min, instead of 12 hours.
2003-09-14 25728, 2003
yalaforge
we were right :-)
2003-09-14 25743, 2003
yalaforge
how do you want to extract the index?
2003-09-14 25755, 2003
djce
Archive::Tar (perl). Or a little variant of it wot I just writ :-)