Requests to Linux development community

[pinecloud]

4 Major Requests to Linux and GRUB development community

1.Give us more choices where to put GRUB/LILO/SILO - should never be defaulted on MBR alone
2.Make GRUB Stage 2 script. ‘menu.lst’ more consistent with true device address
3.Linux device file system mapping to enable over 16 partitions per drive
4.Turn off LVM setup by default unless user ask for it in setup.

My problem happened with open SUSE. It is at most tastefully illustrated with top notch artistic design and of most user friendly package that I admire the painstaking efforts of development communities but it comes with some serious oversight in its boot loader setup. Even if it had such optional setting button somewhere, they are not easily accessible in order to completely disable MBR write and LVM build. Mishap can not be avoided on some beginner like myself. Meaning some damage occur to pre-existing boot loader and partitioning scheme. I have not worked with Ubuntu but Linux developers certainly should listen to our voice.

Of course, I must admit that the GRUB is very intelligent as far as finding root partition where kernel resides with support of some basic Unix command like “find” since it can read nearly all the variants of ufs but a serious shortcoming comes from it’s block device nomenclature which are way too simple in the age of multiple block device interface including SATA, FC, USB, 1394 and like beside historic ANSI ST412/ST506 descendants (16 bit wide MFM, ESDI, IDE and ATA group) and ANSI SCSI (8 and 16 bit).

I have totally been ignorant and never known anything but SCSI until a few month ago. SCSI was so comfortable due to it’s manual jumper setting nature along with dedicated BIOS setup in real mode access per HBA basis to the end.

SCSI and Unix worked well together except some Unix ignored more than 4 physical drives and also multiple HBA (e.g., earlier SCO). Other than that it has always been trouble free. But the production of legacy SCSI drives are coming to a halt. I am stocking enough units of familiar and hardy ST3146855LW spares for the next 5 year usage. These SCSI still outperform SATA II WD Velociraptors at the expense of space and thermal factor.

I am just a graphic artist and photographer a far from Unix IT engineer. However I have noticed undisciplined trends and difference in device enumeration methods amongst different operating system. Once you embed 446 byte initial codes on to MBR or PBR, grub blindly points to what is written in menu.lst text file. If this text file had enough device address information in it such as c0t0d0p0:b (controller ID, target ID, disk ID, partition ID, slice ID and like) instead of just (hd0,0,b) then everything would have worked. GRUB even understands hardware device mapping convention ‘hard-wired’ in the order of historic** or cultural*** hierarchy tradition. That is why when you add a ATA drive on SATA system, GRUB refused to boot because text written device map no loner matches what GRUB see the in the reality. ATA becomes hd0 whist SATA or SCSI become secondary group may be of hd1. I further discovered that GRUB is intelligent enough to remap/swap the device hierarchy or even hide device specified by script and retains it in memory until OS kernel boots but it simply goes by pre-written text script without applying modified environment.

Note**** I assume that “ST”506 means “Schugart Technology” after the name of physicist Alan Schugart (1930 – 2006) who started out as an IBM engineering manager inventing disc drives together with his team, also invented SCSI, then founded Seagate Technology, and later build restaurant business rated with 5 stars and became a writer.

I would like to see someone come up with something like “Joint Unix Device File System Standardisation Committee” together with boot loader designer, Unix developer and chipset manufacturer.

Current GBUB is not very user friendly and forcing it to be on MBR can be harmful. Further it fails to read ext2 or ext3 when BRUB resides on usf2 and fetched from ufs environment.

Such committee also should standardise devfs (device file system) currently having fixed and prewritten 16 device mapping within a disk drive e.g., sd (disk itself) sd1 – sd15 makes total 16 in the age of 1 terabyte drive. NTFS become faster when devided into multiple 64GB partitions. NTFS is not a very intelligent file system as it is advertised. Its bitmap and mft (like giant fat split up to two compornents) and other metafile structure for added extended attributes have been revised without logical order along with it’s updated ifs (installable fs driver). But disc write method is still same as early 1990 era nothing more than RAW version of OS/2 HPFS. In fact HPFS was much more fault tolerant and efficient since its superblock aligned both directions from the mid point of partition. You do pay a price in term of file access speed with a large NTFS partition. Drawback of NTFS is well covered up by successive advancement of disc technology, faster CPU and increased memory. In NTFS, real mft updates do not occur until idle state or sometimes until system shutdown.

Currently some linux installation aborts and refuses to perform disk write when it sees partition count per disk exceeding 16. e.g., Fodora 9. Another distribution makes unsuccessful attempt to consolidate pre-existing partitions to logical volume and leaves some damages behind e.g., SUSE. Whilst others do not get affected and installs correctly but can not access above sda15 which is quite O.K. e.g., Debian. Or else ignore extended partition and go ahead without even looking at logical partitions e.g., Solaris and most of Unix. That’s an admirable trait better than just O.K. I only create and auto-mount a single common vfatfs access point in /mnt and mount pcfs at /share or /export/share with user,rw,exec access anyway.

Currently, since SATA & USB use through ASPI layer for device enumeration, they are assigned as SCSI block device but it will have problem with the presence of true SCSI. If you have combination of ATA, SCSI, SATA and USB, device nomenclature and hierarchy may appear different amongst Linux themselves. Yes, Redhat, SUSE, Debian and of course Solaris all assigned different block device map.



Pinecloud,

Troubled System:

Dell Dimension
Intel Dual Core 3 GHz
2GB Memory
775 Socket 945 Chipset
82801GB PCI-E
SATA II two 300GB WD discs
ATA133 one 75GB Barracuda disc & one DVD
geForce 6600GT PCI-E
Soundblaster Audigy

My Other Typical Systems - Trouble Free:

Two Intel Pentium Tualatine 1.4GHz (In real world faster than P4 2GHz)
1GB Memory
One ICP GDT8623RZ/80803/256MB or One Adaptec 39160
4 Seagate ST3146855LW
One Adaptec 29160 only for CD and scanner
1 Plextor PX40i
1 Plextor PX820i
geForce or Matrox AGP
Soundblaster Live

[GNU-Fan]

Hello,

why do you write this here?

I found the best place to get competent help on any software product is to go where the actual code is worked upon. GNU GRUB - GNU Project - Free Software Foundation (FSF)

Write to the mailing list describing one of your problems a time. You will be heard directly by the actual people in charge for the software. Ask politely and they will either take up the suggestion or at least point you to the right solution.

EDIT: It seems your hassles are more about how the distribution configures/installs the software in question. Then you should talk directly to the persons in charge of the installer of your operating system, of course.

[daark.child]

Pinecloud, you already have several threads discussing the issues above. Can you please stop opening multiple threads and posting more or less the same rant.