GRUB2 - testing report, hppa support?

GRUB2 - testing report, hppa support?

[Nico -telmich- Schottelius]

[-- Attachment #1: Type: text/plain, Size: 1019 bytes --]

Hello!

I am testing grub2 on Soerkis net4801 (x86, geode, serial only, booting
from flash).

The documentation is pretty bad and grub-install is half broken
(will perhaps fix it, when I've time).

grub-setup's --help is confusing, it does not tell the user that
the images will be changed. 

Serial console works at 38400 (others untested).

Though, reprinting the screen / drawing characters is pretty slow,
backspace does not work (using picocom).

You can find my log at
   http://home.schottelius.org/~nico/linux/debug/grub2/grub2-zwerg

Oh, btw, it's HIGHLY confusing that disks start at 0, partitions at
1. Could you please fix it and make it consequently? either hd1,1
or hd0,0, but not hd0,1 or hd1,0.

And I would be interested, whether there's some effort to port
grub2 to hppa? I don't like palo very much I must confess.

Sincerly

Nico

-- 
``...if there's one thing about Linux users, they're do-ers, not whiners.''
(A quotation of Andy Patrizio I completely agree with)

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: GRUB2 - testing report, hppa support?

[Yoshinori K. Okuji]

On Monday 04 December 2006 15:45, Nico -telmich- Schottelius wrote:
> The documentation is pretty bad and grub-install is half broken
> (will perhaps fix it, when I've time).

Any patch is welcome.

>
> grub-setup's --help is confusing, it does not tell the user that
> the images will be changed.

Huh? It says:

Set up images to boot from DEVICE.

If the images wouldn't change, what would you expect from grub-setup?

> Serial console works at 38400 (others untested).
>
> Though, reprinting the screen / drawing characters is pretty slow,

This part requires some optimization. I guess the current implementation of 
the menu interface is not very friendly to a serial terminal.

> backspace does not work (using picocom).

As far as I see the code, there is no reason that backspace doesn't work... So 
this might mean that your terminal sends an unsual escape sequence 
corresponding to backspace.

> Oh, btw, it's HIGHLY confusing that disks start at 0, partitions at
> 1. Could you please fix it and make it consequently? either hd1,1
> or hd0,0, but not hd0,1 or hd1,0.

No. It is consistent with most operating systems, so less confusing to the 
user. GRUB Legacy used 0-based counting for partitions, and I have received 
an uncountable number of complaints. Thus it is really a bad idea to make 
GRUB inconsistent against other systems.

> And I would be interested, whether there's some effort to port
> grub2 to hppa? I don't like palo very much I must confess.

Jeff did something, but I don't know the status.

Thanks,
Okuji

Re: disk vs partition numbering

[Hollis Blanchard]

On Tue, 2006-12-05 at 20:46 +0100, Yoshinori K. Okuji wrote:
> 
> > Oh, btw, it's HIGHLY confusing that disks start at 0, partitions at
> > 1. Could you please fix it and make it consequently? either hd1,1
> > or hd0,0, but not hd0,1 or hd1,0.
> 
> No. It is consistent with most operating systems, so less confusing to
> the user. GRUB Legacy used 0-based counting for partitions, and I have
> received an uncountable number of complaints. Thus it is really a bad
> idea to make GRUB inconsistent against other systems.

I am very glad partitions are now numbered from one; that will certainly
reduce user confusion (although we should expect complaints from people
who know how grub1 works).

However, we will now receive lots of complaints, like this one, because
disks start at zero but partitions start at one. After all, we've all
learned that consistency is critical for good user interface...

-Hollis

Re: GRUB2 - testing report, hppa support?

[Vincent Pelletier]

[-- Attachment #1: Type: text/plain, Size: 1181 bytes --]

Le mardi 05 décembre 2006 20:46, Yoshinori K. Okuji a écrit :
> > Oh, btw, it's HIGHLY confusing that disks start at 0, partitions at
> > 1. Could you please fix it and make it consequently? either hd1,1
> > or hd0,0, but not hd0,1 or hd1,0.
>
> No. It is consistent with most operating systems, so less confusing to the
> user. GRUB Legacy used 0-based counting for partitions, and I have received
> an uncountable number of complaints. Thus it is really a bad idea to make
> GRUB inconsistent against other systems.

But by changing the numbering, we void both the learning effort of the 
people "who knows" and the answering effort they spent with people "who don't 
know". At least, that's my position regarding to the change.

By changing it we might both face the complaints of the people - like me :) - 
who think the two numberings are inconsistent and the people who would have 
anyway complained about starting from 0, because disks are still numbered 
from 0... We may also face complaints from the people who already complained 
about starting from 0 - and eventually accepted it - and who will have to 
learn again now...

-- 
Vincent Pelletier

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: GRUB2 - testing report, hppa support?

[Jan C. Kleinsorge]

Vincent Pelletier wrote:
> Le mardi 05 décembre 2006 20:46, Yoshinori K. Okuji a écrit :
>   
>>> Oh, btw, it's HIGHLY confusing that disks start at 0, partitions at
>>> 1. Could you please fix it and make it consequently? either hd1,1
>>> or hd0,0, but not hd0,1 or hd1,0.
>>>       
>> No. It is consistent with most operating systems, so less confusing to the
>> user. GRUB Legacy used 0-based counting for partitions, and I have received
>> an uncountable number of complaints. Thus it is really a bad idea to make
>> GRUB inconsistent against other systems.
>>     
>
> But by changing the numbering, we void both the learning effort of the 
> people "who knows" and the answering effort they spent with people "who don't 
> know". At least, that's my position regarding to the change.
>
> By changing it we might both face the complaints of the people - like me :) - 
> who think the two numberings are inconsistent and the people who would have 
> anyway complained about starting from 0, because disks are still numbered 
> >from 0... We may also face complaints from the people who already complained 
> about starting from 0 - and eventually accepted it - and who will have to 
> learn again now...
>
>   
> ------------------------------------------------------------------------
>
> _______________________________________________
> Grub-devel mailing list
> Grub-devel@gnu.org
> http://lists.gnu.org/mailman/listinfo/grub-devel
>   
An escape from this 'felt' inconsistency could be to index the drives
with alphabetic characters instead.

-Jan

Re: disk vs partition numbering

[Yoshinori K. Okuji]

On Saturday 09 December 2006 01:17, Hollis Blanchard wrote:
> On Tue, 2006-12-05 at 20:46 +0100, Yoshinori K. Okuji wrote:
> > > Oh, btw, it's HIGHLY confusing that disks start at 0, partitions at
> > > 1. Could you please fix it and make it consequently? either hd1,1
> > > or hd0,0, but not hd0,1 or hd1,0.
> >
> > No. It is consistent with most operating systems, so less confusing to
> > the user. GRUB Legacy used 0-based counting for partitions, and I have
> > received an uncountable number of complaints. Thus it is really a bad
> > idea to make GRUB inconsistent against other systems.
>
> I am very glad partitions are now numbered from one; that will certainly
> reduce user confusion (although we should expect complaints from people
> who know how grub1 works).
>
> However, we will now receive lots of complaints, like this one, because
> disks start at zero but partitions start at one. After all, we've all
> learned that consistency is critical for good user interface...

Agreed. Yes, consistency is extremely important. Unfortunately, there are two 
types of consistency in this world:

- Mathematical or symmetrical consistency

- Customary or accustomed consistency

As you know very well, GRUB Legacy follows the former. I decided to change it 
to the latter in GRUB 2, as I don't have to care about compatibilities with 
GRUB Legacy so much, and I learned that theoretical beauty is often just a 
masturbation when coming to the user interfaces with experience.

There are so many "inconsistencies" in computers. For example, lines are 
counted from 1. Columns are counted from 0. AFAIK, all editors and viewers 
follow this convention. If one makes it "consistent", probably a lot of 
people would feel uncomfortable.

The critical thing is how to reduce new things that people would have to study 
for using a program. GRUB Legacy made a mistake, since nearly all operating 
systems use 0-based for disks, and 1-based for partitions. In GRUB Legacy, 
compatibilities preceded the learning curve. That is why one chapter is used 
only to train people for a new thing in the manual! The same thing is 
described more formerly in a later chapter as well. This is the cost in 
having inconsistency with other materials. Not about the cost in writing 
documentation. It is about the cost in millions of people reading many lines 
again and again.

Now, some people say that this is inconsistent against GRUB Legacy. OK. I 
admit it. But which is more important in a long run: easy for existing users 
to migrate to GRUB 2, or easy for new comers to adapt GRUB 2? How difficult 
is it that existing users know GRUB now follows the same rule as others? How 
difficult is it that beginners study a rule different from others, so not 
intuitive at all?

Okuji

Re: disk vs partition numbering

[Hollis Blanchard]

On Tue, 2006-12-12 at 23:46 +0100, Yoshinori K. Okuji wrote:
> 
> The critical thing is how to reduce new things that people would have to study 
> for using a program. GRUB Legacy made a mistake, since nearly all operating 
> systems use 0-based for disks, and 1-based for partitions.

Since I'm used to Linux, which assigns disks letters rather than
numbers, I was unaware that other OSs like BSD number disks from 0.
Since that's the case, and neither 0 nor 1 will make sense to people
expecting "a", I'm OK with 0.

> Now, some people say that this is inconsistent against GRUB Legacy. OK. I 
> admit it. But which is more important in a long run: easy for existing users 
> to migrate to GRUB 2, or easy for new comers to adapt GRUB 2? How difficult 
> is it that existing users know GRUB now follows the same rule as others? How 
> difficult is it that beginners study a rule different from others, so not 
> intuitive at all?

Yes, I definitely agree with the decision to number partitions from 1.
(Hopefully we can do the same with menu entries too. :) Making this
change at a time when the config file syntax changes is the only good
opportunity.

-Hollis

Re: disk vs partition numbering

[James Lockie]

Yoshinori K. Okuji wrote:
> On Saturday 09 December 2006 01:17, Hollis Blanchard wrote:
>   
>> On Tue, 2006-12-05 at 20:46 +0100, Yoshinori K. Okuji wrote:
>>     
>>>> Oh, btw, it's HIGHLY confusing that disks start at 0, partitions at
>>>> 1. Could you please fix it and make it consequently? either hd1,1
>>>> or hd0,0, but not hd0,1 or hd1,0.
>>>>         
>>> No. It is consistent with most operating systems, so less confusing to
>>> the user. GRUB Legacy used 0-based counting for partitions, and I have
>>> received an uncountable number of complaints. Thus it is really a bad
>>> idea to make GRUB inconsistent against other systems.
>>>       
>> I am very glad partitions are now numbered from one; that will certainly
>> reduce user confusion (although we should expect complaints from people
>> who know how grub1 works).
>>
>> However, we will now receive lots of complaints, like this one, because
>> disks start at zero but partitions start at one. After all, we've all
>> learned that consistency is critical for good user interface...
>>     
>
> Agreed. Yes, consistency is extremely important. Unfortunately, there are two 
> types of consistency in this world:
>
> - Mathematical or symmetrical consistency
>
> - Customary or accustomed consistency
>
> As you know very well, GRUB Legacy follows the former. I decided to change it 
> to the latter in GRUB 2, as I don't have to care about compatibilities with 
> GRUB Legacy so much, and I learned that theoretical beauty is often just a 
> masturbation when coming to the user interfaces with experience.
>
> There are so many "inconsistencies" in computers. For example, lines are 
> counted from 1. Columns are counted from 0. AFAIK, all editors and viewers 
> follow this convention. If one makes it "consistent", probably a lot of 
> people would feel uncomfortable.
>
> The critical thing is how to reduce new things that people would have to study 
> for using a program. GRUB Legacy made a mistake, since nearly all operating 
> systems use 0-based for disks, and 1-based for partitions. In GRUB Legacy, 
> compatibilities preceded the learning curve. That is why one chapter is used 
> only to train people for a new thing in the manual! The same thing is 
> described more formerly in a later chapter as well. This is the cost in 
> having inconsistency with other materials. Not about the cost in writing 
> documentation. It is about the cost in millions of people reading many lines 
> again and again.
>
> Now, some people say that this is inconsistent against GRUB Legacy. OK. I 
> admit it. But which is more important in a long run: easy for existing users 
> to migrate to GRUB 2, or easy for new comers to adapt GRUB 2? How difficult 
> is it that existing users know GRUB now follows the same rule as others? How 
> difficult is it that beginners study a rule different from others, so not 
> intuitive at all?
>
> Okuji

I like that fact that grub is different. :-)
All software should refuse to work unless the manual is read, grub 
legacy was like that for me. :-)
I take that back, software should work without having to read the 
manual, grub legacy was NOT like that for me.
Maybe grub should use the way of designating a disk/partition in the 
same way as what OS it is running on?
Yes, being different for each OS. :-)

Re: GRUB2 - testing report, hppa support?

[Marco Gerards]

Vincent Pelletier <subdino2004@yahoo.fr> writes:

> Le mardi 05 décembre 2006 20:46, Yoshinori K. Okuji a écrit :
>> > Oh, btw, it's HIGHLY confusing that disks start at 0, partitions at
>> > 1. Could you please fix it and make it consequently? either hd1,1
>> > or hd0,0, but not hd0,1 or hd1,0.
>>
>> No. It is consistent with most operating systems, so less confusing to the
>> user. GRUB Legacy used 0-based counting for partitions, and I have received
>> an uncountable number of complaints. Thus it is really a bad idea to make
>> GRUB inconsistent against other systems.
>
> But by changing the numbering, we void both the learning effort of the 
> people "who knows" and the answering effort they spent with people "who don't 
> know". At least, that's my position regarding to the change.
>
> By changing it we might both face the complaints of the people - like me :) - 
> who think the two numberings are inconsistent and the people who would have 
> anyway complained about starting from 0, because disks are still numbered 
> from 0... We may also face complaints from the people who already complained 
> about starting from 0 - and eventually accepted it - and who will have to 
> learn again now...

You can never make everyone happy.  Whatever you propose, there will
always be someone who disagrees.

--
Marco

Re: disk vs partition numbering

[Yoshinori K. Okuji]

On Wednesday 13 December 2006 05:11, Hollis Blanchard wrote:
> Since I'm used to Linux, which assigns disks letters rather than
> numbers, I was unaware that other OSs like BSD number disks from 0.
> Since that's the case, and neither 0 nor 1 will make sense to people
> expecting "a", I'm OK with 0.

Not always true. For example, look at /dev/fd* and /dev/ide/hd/*. Linux is 
inconsistent, but does use 0-based counting for disks when using integers.

> Yes, I definitely agree with the decision to number partitions from 1.
> (Hopefully we can do the same with menu entries too. :) Making this
> change at a time when the config file syntax changes is the only good
> opportunity.

I didn't think of this. It is an excellent question. For now, I don't know 
which is more natural, starting from 0 or 1 for menu entries. Marco, what is 
your opinion about this? Since you wrote most of the code, I guess you have 
some suggestion.

Okuji

Re: disk vs partition numbering

[Yoshinori K. Okuji]

On Wednesday 13 December 2006 09:14, James Lockie wrote:
> I take that back, software should work without having to read the
> manual, grub legacy was NOT like that for me.

Exactly. It is ideal that one can use software without reading a manual 
appropriately. The reality is different, but we should make effort to get 
closer to this goal.

> Maybe grub should use the way of designating a disk/partition in the
> same way as what OS it is running on?
> Yes, being different for each OS. :-)

Unfortunately, this would cause strange behaviors. For example, suppose that 
you have installed GRUB into a floppy or CD-R. When you bring it to another 
computer, how should GRUB behave? Mimic the operating system on which GRUB 
was installed? Mimic the operating system the user is trying to boot? If you 
have multiple operating systems installed, which one???

Okuji

Re: disk vs partition numbering

[Tomáš Ebenlendr]

Dne 13 Prosinec 2006, 22:00, Yoshinori K. Okuji napsal(a):
> On Wednesday 13 December 2006 09:14, James Lockie wrote:
>
>> I take that back, software should work without having to read the
>> manual, grub legacy was NOT like that for me.
>
> Exactly. It is ideal that one can use software without reading a manual
> appropriately. The reality is different, but we should make effort to get
> closer to this goal.
>
>> Maybe grub should use the way of designating a disk/partition in the
>> same way as what OS it is running on? Yes, being different for each OS.
>> :-)
>>
>
> Unfortunately, this would cause strange behaviors. For example, suppose
> that you have installed GRUB into a floppy or CD-R. When you bring it to
> another computer, how should GRUB behave? Mimic the operating system on
> which GRUB was installed? Mimic the operating system the user is trying to
> boot? If you have multiple operating systems installed, which one???
>
> Okuji

The only think grub may (should?) mimic is bios. IBM-PC compatible bios
has no 'visible' numbering. I would choose 0-based, (subtracting 0x80
is better for me than 0x81), but I'm not a beginner.

Another thing is, that we can add possibility to change naming scheme
by some module. Then, at the beginning of config file will be something
like 'renamedevices --linuxstyle', and then we can use (hda,1).


--
                            Tomas Ebenlendr
                            http://drak.ucw.cz/ebik