Darren J Moffat wrote: > Pavan Chandrashekar - Sun Microsystems wrote: > >> Darren J Moffat wrote: >> >>> Moinak Ghosh wrote: >>> >>>>> Yes but the partition IDs that this project is suggesting be used >>>>> are explicitly tagged as being for DOS or Windows. >>>> >>>> >> >> Ok. >> Extended partitions is not new to OpenSolaris. PCFS handles it >> internally. > > > That is completely different. The current support is ONLY for reading > DOS/Windows (ie FAT based) filesystems. > >> It is convenient to have them supported in OpenSolaris. Most other >> OSes support it. It is part of the fdisk partitioning scheme. > > > I have no objection to having OpenSolaris (or Solaris for that matter) > support a properly layered partitioning system so that the knowledge > of what is a "primary" vs an "extended" DOS/Windows partition is not > part of the filesystem driver. > > No objection what so ever. > > However that is NOT how you are presenting this project, it is being > presented as having the ability to install an OpenSolaris distro into > a DOS/Windows fdisk partition. > > If we look at the orginal proposal again: > >> The aim of the project is to enable the installation and booting >> of OpenSolaris from an extended partition. > > > This is the bit I have issue with. > >> This project will be delivered in multiple phases. The first phase is >> to introduce all OS changes necessary to support booting from and >> managing extended partitions on OpenSolaris. Here are the list of >> changes for phase 1 : >> >> 1. Driver Changes ( cmlb : modification) >> 2. Tools to perform partitioning (fdisk, format : modification) >> 3. Library to provide partitioning support (libfdisk : new) > > > I have no really issue with any of that providing: > > --->> a) this is SPARC and x86 <<---
Darren, so far so good. But isn't your point a bit ridiculous here (plus unfair and harming a promising project, that would undoubtedly remove a barrier for certain "aliens" to give OpenSolaris a very first try on their x86/x64 box[es]) ? Striving to keep the different ISAs in sync is excellent, but one has to stay realistic. Don't forget the opposite byte orders: sparc MSB x86/x64 LSB pccle LSB (2.5.1 and 2.6beta) ppc MSB (In 1995 Sun-Labs internal bige. 2.6 study / now the upcoming Polaris) For example, can you mount sparc-ufs on x86/x64 ?? --->> NO Vise versa --->> NO Not even readonly. Only ZFS can be moved. So where is the compatibility you are calling for? Okay, I was talking about filesystems, not partitions. But I also never saw a sparc system recognizing a x86's VTOC, nor the other way around (correct me if I'm wrong). (same for 2.5.1ppc ! ) If we are talking about data structures like linked lists here, byte order differences slightly complicate the situation. Only hdds containing a primary partition with fat12, fat16 or fat32 can be mounted on a SPARC (via the pcfs driver that properly interprets the partition table / no VTOC of course). The other question is, admit it: Who demands the functionality offered by the proposed "OpenSolaris on extended partitions" project on SPARC ?? ### I welcome Pavan's project, because I do know, that most x86-folks out there are either using WinNT or LinUX, unfortunately :-( (Most of the time _both_, but not Solaris_x86.) Just read this: Sun Microsystems, the begrudging Linux vendor http://searchopensource.techtarget.com/originalContent/0,289142,sid39_gci1238478,00.html Solaris has to be more flexible in order to bring outsiders to giving it a first try: linUX can even be installed into a plain WinDOwS9x file! 100 people may do that, and 5 of them might completely switch over after having had their first linUX experience. One out of 100 is now a valuable kernel coder and potentially contributes a lot. Why would someone be against that so called "user friendliness"? Being against this proposal means actively harming Solaris' prevalence. -- Martin Boochnig > b) This results in knowledge of partitioning being removed from pcfs > c) The project aims to define how we name thinks in /dev/dsk as a > result of this new knowledge of "extended" partitions. > >> 4. Grub changes ( GRUB and installgrub : modification) > > > This looks like it is only needed for install. > > Basically what I'm saying is that for me the real problem is that pcfs > is the place today that knows about extended partitions that needs to > change it needs to be in a more appropriate place. I agree that we > need thinks like fdisk/format etc updated. We also need to make sure > that thinks like the Tamrack project get to benefit from this. > > > It is as much windows/dos specific as much as a primary fdisk > partition is. > > I disagree, if that were actually true there would be only two types > of partition "primary" and "extended" that isn't the case there are > many many tags for partitions. Linux would not have needed a "Linux > extended" partition type that is different to the "DOS/Windows" or > "Windows 95" ones. > >> It wont raise special cases for co-existence with other OSes like >> Linux. There is no specified limit on the number of logical drives in >> the > > > Logical drives ? A DOS/Windows term :-) > >> extended partition. As long as there is space on the disk and the >> disk space is covered by the extended partition, you can grow the >> number of >> logical drives. > > > The very fact you talk about logical drives rather than about putting > a VTOC in place and defining slices betrays this as DOS/Windows > functionality. > > _______________________________________________ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
