[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: UUID - autmatically entries? (?OT)

On Sat, 21 May 2011 13:51:27 -0600, Paul E Condon wrote:

> On 20110521_175742, Camaleón wrote:
>> I'd say there is no perfect method to manage this.
> But sometimes the perfect is the enemy of the vastly better. I simply
> cannot believe that the current situation will persist through the next
> several releases of Debian. Something better will be found, IMHO.

Sure, sometimes what the vast majority think is better is not the 
default :-)

But the current situation is here to stay, it's an upstream decision that 
was taken years ago and we (in fact, not just we at Debian, but any other 
distribution) have to stick to it. 

It could be somehow improved, I agree, but finding a universal and 
uniquely way of doing things (and that works) is not that easy :-)
>> I still miss the old-plain-intuitive method for designating block
>> devices (hda → first ide, hdb → second ide device, sda → first
>> scsi/sata device, etc...) but we have to cope the new changes because
>> they are vry much needed (nowadays we connect many devices of diverse
>> nature, most of them hot-swappable and they cannot overlap).
>> In this regard, every user chooses the best method for identifying hard
>> disks (I like labels for removable devices and uuid/id for fixed hard
>> disks) but lastly this is just a matter of convenience, here there is
>> no "one size fits all", no method is perfect for all the situations and
>> all of them have their own drawbacks.
> I agree that labels are good on hot pluggable devices. I think that
> happens under Gnome, at least in some circumstances. (or at least it has
> happened on my computer at times when I have had Gnome installed.)
> I think labels could be made to work for internal devices with a small
> measure of cooperation from the installer. What is needed, in my naive
> opinion, is that the installer use a label if one present on a partition
> and if it does not conflict with a label that it has already seen in a
> once through scan of the block special devices. Only if there is a name
> collision would it use a UUID, and it would generate a new UUID only if
> the pre-existing UUID conflicts with an already seen the UUID that is on
> a unlabeled partition.

There can be many problems with "labels", and having problems with the 
internal storage can lead to severe and unexpected results I really 
prefer to avoid, and for that reason (and for internal storage) I'd go 
for "uuid" or "id" as the default. Yes, that's not a human readable 
format but in the end, is the computer which has to deal with it, not 
us... we could masquerade those lengthy names in the same way we use a 
DNS server for domain names to find/translate the numeric IP address of 
billions machines in the world :-)



Reply to: