[I apologize if this duplicates an earlier message; I know I lost a message
when the Mac I was telnetting from crashed, and I think it was the one where
I raised this idea, but I'm not certain.  If this is a duplicate, I think I
left out the explanation I was going to provide...]

> > I hope any CDROM makers will try their best to include all the
> > sources in non-compressed form!

>  You know what this means, do you? Thousand of small files, most of them
> smaller than 2048 Bytes Chunks.

There are about 20000 files on my /usr/src partition, which is all the MI
sources plus the sys/arch/i386 tree.  That's not going to overflow a CDROM
all by itself because of block alignment.  (The average file size seems to
be around 10K, but that might be dominated by a few very large files in
/usr/src/share.)

Someone mentioned the possibility of having a tar filesystem type; given the
ability to do that, it wouldn't be too hard to have a gzipped-tar filesystem
type as well, which would allow sources to be efficiently stored on the CDROM
yet easily browsable just like a real filesystem.  You'd probably want a mutant
form of gzipped tar, though:  instead of gzipping an entire tar archive, you'd
probably want to tar up gzipped files instead.  You could speed repetitive
searches somewhat by employing various caching strategies; caching file names
and offsets is one obvious strategy, caching actual extracted files if
referenced more than once is another.  (Of course, if you're running with a
cramped ramdisk root and a CDROM image, the caching isn't going to help
much...)