On 23 March 2014 21:56, lee <lee(a)yun.yagibdah.de> wrote:
There`s nothing weird or exotic about it. I`ve always had /usr on
its
own partition until the F17 installer refused that, which it shouldn`t
have.
As I have commented elsewhere, I think this is a 1980s style of
thinking. Things have changed. Move on. Sorry, but they have; see the
links elsewhere.
RAID isn`t exotic, either. Disks do fail, the only question is
when,
and I neither want to lose data, nor the hassle.
Sure, I use it on all my servers.
Installing on a laptop requires encrypted partitions. They can be
stolen too easily.
I have never ever used this and never expect or plan to. I suggest
that your blanket statement is too sweeping.
http://xkcd.com/538/
> I never separate out /tmp or /var or /usr/local - I only ever use
/
> and /home basically.
I always use separate partitions. It has lots of advantages.
As I said, I use / and /home and advise against combining them.
Personally I think that's enough. I am not disputing your reasons, but
AIUI, Fedora is trying out a move to flatten and simplify the
way-too-complex directory hierarchy. It's happened. The decision is
made. Deal with it, move on.
/var can get full, and it`s written to, same goes for /tmp.
As I said elsewhere: when the smallest new HD you can buy is half a
terabyte (and even SSDs start at half that) this really isn't a big
issue any more.
How do you
mount /usr read-only?
I don't. Never have in 26y of Unix systems support. For rescue, now, I
boot off a LiveDVD or LiveUSB.
Especially on a server, it`s a good idea to mount
everything read-only that you can. When you have several disks, you can
do your partitioning in such a way that you get better performance.
An illustrative anecdote, for what it's worth:
Back in 1995 I was the testing labs manager for a leading UK computer
magazine. In an article, I recommended partitioning 1.2GB disks with a
dedicated swap partition at the end. This meant the main partition
could be under 1GB and thus use far more space-efficient 8kB blocks.
(This is on Win95 or NT 3.5, before FAT32 was invented.)
A system manufacturer complained that putting swap on the end of the
drive would kill performance. I disagreed. He complained more.
So he came in and I set up a test and showed him that, to 2 decimal
places in a percentage-based benchmark score, i.e. well below
measurement error, there was absolutely *no* difference between the
speeds of different areas of the disk.
This is in 1995, when LBA addressing was new, 1.2GB was a big disk,
and DMA was just starting to appear for EIDE drives.
Back then, the difference was not measurable.
That was nearly 20y ago.
Now, it is not real. It is not there any more. Disks are a thousand
times bigger and faster now. This stuff does not matter any more and
hasn't since before Linux 2.0 was released.
Especially when you have a server, you may need a (pretty much)
granted
capacity on /var or /tmp to make sure it will continue to operate ---
without separate partitions, your users may fill up the disks ...
Users shouldn't be able to write stuff to / at all! Only to /home or
below, or dedicated data partitions. Where /usr or /var is should not
matter to them.
Nowadays you may have SSDs which supposedly last longer when not
written
much to but mostly read from, so you might put the partitions that can
be read-only on the SSDs and use magnetic disks for things like /var,
/tmp, /home and swap.
Machines come with dozens of gigs of RAM now. I'm not sure there's
much argument for swap at all, and personally, I use tmpfs for better
performance and a self-cleaning /tmp tree.
Why wouldn`t you use different partitions? I can see it (and have
done
it) for when the available disk capacity is extremely limited, but
otherwise it doesn`t make any sense and has nothing but disadvantages.
Exactly. One splits stuff up when space is an issue. When it isn't,
one doesn't need to.
It`s merely a reasonable standard thing to use separate partitions
and a
requirement to use RAID, and encrypted partitions for laptops, not
something in any way unusual. Of course I expect an installer to handle
that as well as using a single, unencrypted partition on a single disk.
Actually, I agree, but there must be /some/ limits to the granularity!
And it`s not too difficult. The installer doesn`t need to do the
partitioning, the user does it. The installer only needs to give the
user a good tool to do the partitioning the user wants and let them use
it. Good tools to do partitioning are already available, and the
installer doesn`t need to re-invent the wheel in that.
I mostly agree, but bear in mind that the installer must cope with
both experts and novices. That's a tough call.
I'd say it's /too/ simplified at the moment, though, and I think you
might agree...?
Perhaps it even shouldn`t. Why force the user to learn how to use
yet
another partitioning tool they even rarely use unless they install
Fedora all the time? Why not give them a choice, like either cfdisk or
parted, then tell the installer what to do with each partition and let
them switch between these until they are done --- or let the installer
do whatever partitioning it wants, which means that all existing data on
the disks will be deleted.
Mostly, I would agree, actually.
--
Liam Proven * Profile:
http://lproven.livejournal.com/profile
Email: lproven(a)cix.co.uk * GMail/G+/Twitter/Flickr/Facebook: lproven
MSN: lproven(a)hotmail.com * Skype/AIM/Yahoo/LinkedIn: liamproven
Tel: +44 20-8685-0498 * Cell: +44 7939-087884