I do not want to roll back 40 years of progress.  Actually, I was 
thinking about this last night.  I started programming 47 years ago.  My 
first system was one where I learned to punch binary on an 80 column 
card.  The mass storage device was a card punch.  We learned how to use 
a card sorter.  Later, things improved to where I was able to use a 
paper based line printer and punched tape.  What a joy it was in the 
70's to use a 'glass tty' where you could edit one line at a time.

My first high level language was FORTRAN.  Not Fortran II or FORTRAN IV. 
The original FORTRAN.  My wife says she should have know better when out 
'dates' consisted of going to the computer center to wait for output.

My first 'home' computer cost $2000 and I had to put it together with a 
soldering iron.  I remember buying 16K RAM for $150 and my first 80M 
hard drive for the bargain price of $600.  Why that's less than $1 per M!

My, how things have changed!

The problem with the current environment is not that we continue to 
progress, but that a very few people are making changes in a very 
autocratic way.  These guys are very smart, but they are making changes 
to fix their relatively exotic problems in a way that affects me when I 
do not have those problems.  For me, it's not broke, so don't force me 
to 'fix' it.

I am not really against combining /bin and /usr/bin, etc.  The reason 
that they were separate in the first place was that disks were small and 
expensive.

Today we have technical problems that force the change.  The MSDOS or 
PC98 standard for the Master Boot Record does not support addresses 
greater than 2T.  We want to boot from all types of devices: IDE Drives, 
SCSI Drives, USB Drives, SATA Drives, CDROM Drives, DVD Drives, Network 
Cards, Partitions formatted as RAID, LVM, EXT2/3/4, ReiserFS, XFS, JFS, 
encrypted drives, etc.  And that's just for Linux and doesn't count 
other OSs.

My problem with the newer systems is that they want to create a 'one 
size fits all' system.  There is only one legitimate reason for an 
initramfs -- to mount the root directory.  This is what the /usr, /bin, 
/sbin, and /lib directories were for.  Now I agree that mounting the 
rootfs in every case is beyond the scope of the kernel, but why does 
every combination have to be supported?  Does /usr really need to be 
encrypted?  Does it *need* to be on an LVM partition?  Use one ext4 
partition for /, one for /boot, and put everything else on an LVM 
partition and the boot process becomes quite straight forward.  No 
initramfs and it's accompanying opaqueness required.

Do we need systemd, a complex compiled system, to boot?  My LFS boots in 
8 seconds now.  Is that too long?  Sure there are other issues, but I 
haven't seen them.  Why is upstream 'fixing' my system that is not broken?

   -- Bruce

-- 
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Reply via email to