Greg Freemyer wrote:
> On 9/30/07, Richard Creighton <[EMAIL PROTECTED]> wrote:
>   
>> Carlos E. R. wrote:
>>     
> <snip>
>   
>>> That's not quite so now. For instance, the limit on the number of
>>> partitions has been decreased from 64 to 16 (less than). That's one of the
>>> consequence of "progress" in the linux field.
>>>
>>> - --
>>> Cheers,
>>>        Carlos E. R.
>>>
>>>       
>> Someone made an ill-advised decision to change the naming scheme of IDE
>> drives to be the same as the new SATA drives to be the same as SCSI.
>> In the process, it inherited the limitations of the SCSI drives.   I
>> can't think of a reason for having done it, but it appears to have been
>> done in all the distros.   I suspect there will be a great gnashing of
>> teeth when the next release hits the streets and some accomodation will
>> be forthcoming.   As one of the beta testers for upcoming 10.3 SuSE, it
>> has already proven 'interesting' and caused me personally no end of
>> frustration.   Generally though, Linux's progress has kept pace with the
>> newer hardware without losing sight of its historical past.  This is one
>> of the few exceptions so far.   I bet that there is NO chance that XP,
>> much less Vista will run on a 386 or a 286...  I cranked up 10.2 on a
>> 486DX-2 the other day just to see it run...slow, but it ran :)
>>
>>
>> Richard
>>     
>
> The issue is far deeper than naming conventions.
>
> When SATA support was added to the kernel (libata) they leveraged the
> entire SCSI subsystem due to its quality compared to the IDE
> subsystem.
>
> Then libata got to so good that many (most) of the PATA drivers were
> re-implemented (by Alan Cox of Redhat) via libata.  And then the new
> implementations got stable enough that the distros decided to move to
> the libata pata drivers by default.  (Fedora was the first to move in
> the spring.)
>
> But, for the foreseeable future you should be able to use the old
> drivers/ide implementation and get the old functionality (and naming
> convention).
>
> The long term solution is to have libata implement its own full set of
> infrastructure and no longer fit under the SCSI infrastructure.  When
> that happens the partition limits should be restored to the higher
> limits.
>
> Novell has Tejun Heo supporting libata.  He has done 2 major upgrades
> to it in the last 18 months (new error handling logic for 10.2, PMP
> support for 10.3, ??? for 11.0).
>
> My hope is that his next big project will be the libata
> infrastructure, but I have not seen anything posted about that yet.
>
> Greg
>   
Greg, I usually do a lot of <snipping> but your reply deserves to be
seen in context and completeness.  

What you say is quite probably accurate.   The problem is that the
implementation was made  in such a way as to cause unexpected problems. 
For many people, it probably was uneventful, but a lot of   people have
mixed IDE and SATA architectures with motherboards that have both types
of controllers.   These motherboards typically offer the IDE drives to
the OS first, then the SATA drives.   I had a perfectly good 10.2
installation on my IDE system drive and added SATA drives to the machine
for testing the 10.3 beta.  I did not know of the unadvertised decision
to incorporate IDE into the SATA/SCSI family.   Properly and COMPLETELY
done, there would have been no problem.    Had there been a warning I
could have taken precautions but from 9.3 through 10.2 when I have
upgraded, I have never needed to worry about such things.   Well, the
install went fine, the 10.3 beta installed onto the SATA drive, but the
IDE drive normally called hdb1 on my system was suddenly renamed sda2
(my cdrom was on the master channel) and the first SATA drive became
sdb1 and when grub was written and the MBR was written, guess where it
was put....yup, it clobbered my IDE drive.   Of course, this was totally
unexpected.   OBTW, I neglected to tell you, the SATA drives were not
just 1 drive, they were 4 200G drives in a MD Raid configuration which
also confused the issue, but that is no reason to clobber the IDE MBR or
rewrite it's /boot or /boot/grub entries.   To this day, if I want to
install any version above 10.3 beta 1, I have to unplug the IDE drive
until the installation is complete.   Also, the YaST repair modules are
totally clueless still about the name changes.  

So, while the motive to migrate is honorable, the decision to do so
unilaterally and without proper notice and warning about possible side
effects between minor releases I maintain is and was still ill
advised.   To do so between version 10 and version 11 would have been
much more appropriate, but from .2 to .3, that usually signifies
relatively minor changes and enhancements and bug fixes and not major
changes.   Since I can recall, IDE devices have been called HDxx and
drivers and software buried pretty deep has expected this for years.  
To suddenly change this invites trouble, and it happened.   I would have
simply unplugged the IDE drive if I had expected trouble, but the
stability of SuSE upgrades in the past even in beta has never put drives
not part of the experiment at risk before and I grew complacent ... my
bad,  add a bad decision on the development end and a lot of   people
are going to wonder what happened.   The buglist reports bear that out
already.

I do appreciate the insight of your reply however, thank you.

Richard

-- 
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to