[rhelv6-beta-list] My first experiences with RHEL6 beta

[Bryan J. Smith]

R P Herrold <herrold at owlriver.com> wrote:
> This misses the point.  It is not a lack of
> 'enlightenment' at all.  LVM is 'ignored' (removed from
> consideration, actually) because it tries to replace a
> robust and understood tool with a perhaps more capable one,
> at the expense of adding complexity to a system.  The
> 'value proposition' fails in readily understood use cases

Just to review, and as I have seen it, there are three (3) different
contexts of this "meta-discussion" being responded under:  

1.  Performance -- which I've been focused on.  Based on the prior
    assumption that "swap file is close to data" for single disk,
    newer kernel and DeviceMapper facilities are 100x more capable
    in this regard, as legacy disk labels cannot for file systems.

I'm going to defer _all_ future commentary on #1 to producing 
documentation, instead of the list.  That's what I was referring to.  ;)

2.  Flexibility -- which I pointed out is impossible if people don't
    follow good LVM practices.  I shouldn't have even gone down that
    road, because it was already a narrow, hypothetical case.

3.  Complexity -- no argument there.  Although even looking at a single
    disk, the legacy MBR partition table / PC BIOS has serious limits
    we're now running into.

Yes, #3 is interesting, because ...

> - The older methods are almost self documenting on a single
> man page;

Not entirely true.  In fact, that -- among other assumptions presented
here -- are not at all, even for a single hard drive.  That's why all,
new versions of major distros have changed to UUID and other references,
and _away_ from legacy, fixed devices.

Furthermore, at this time, one should be referencing /dev/disk and _not_
other references, as they are not guaranateed nor flexible enough.  This
too needs extensive documentation.

As I recently covered in a LUG meeting on Linux Devices and Storage
for the 21st century, today's "single, internal hard drive" systems can
have over a dozen different devices (nearly all external -- basically
everything but ONFI) named into /dev/sd* devices.

This is why I'm going to focus on documenting these details, instead of
focusing on what people remember and assume from the 20th century.  I
wouldn't point these things out if I hadn't seen the issues ...

  on single hard drive desktop PCs no less  ;)

Peace.