[9fans] Fossil defaults (again)

[9fans] Fossil defaults (again)

[tlaronde]

Since I'm playing with the installation procedure (to be able to install
Plan9 without a CD reader; without pxe; without using GRUB+etherboot...:
done... Plan9 has everything once the FAT+PBS+9load interactions are
understood), I will have to modify several install scripts since, for
example, the scripts infer from the presence of a plan9:fat that
everything is OK---while I create a FAT from another system to be able
to bootstrap the install.

And since I'm reviewing things, two details :

1) In disk/prep/prep.c, the default minimal size for fossil is:

    {   "fossil",   200*MB, 0,  4,  },

With a distribution that is more than that, and the snapshot dance,
should this not be bumped?

2) Still in the same src, for swap:

    {   "swap",     100*MB, 512*MB, 1,  },

but pc/inst/fmtfossil sets from the definition of the swap the
value for fossil ncache. If the minimal default for swap will lead
to:

         m=256   # 2MB for fossil (this will be slow)

wouldn't it be better to increase the minimal size of swap?

--
        Thierry Laronde <tlaronde +AT+ polynum +dot+ com>
                      http://www.kergis.com/
Key fingerprint = 0FF7 E906 FBAF FE95 FD89  250D 52B1 AE95 6006 F40C

Re: [9fans] Fossil defaults (again)

[erik quanstrom]

> 1) In disk/prep/prep.c, the default minimal size for fossil is:
>
>     {   "fossil",   200*MB, 0,  4,  },
>
> With a distribution that is more than that, and the snapshot dance,
> should this not be bumped?

this only applies to automatic partitioning.  if you set up
partitions manually, it doesn't apply.  what do you think
automatic partitioning should do?  should it only be of use
for installing the distribution, or should it be of general use?

if your answer is the former, then prep is pretending to
be general, when it's not.  and 200mb might be too small.

if your answer is the latter, then 200mb might be too large,
it might make sense to sometimes set up a very small fossil
for stand-alone servers.  say, a stand-alone secstore server.

> 2) Still in the same src, for swap:
>
>     {   "swap",     100*MB, 512*MB, 1,  },
>
> but pc/inst/fmtfossil sets from the definition of the swap the
> value for fossil ncache. If the minimal default for swap will lead
> to:
>
>          m=256   # 2MB for fossil (this will be slow)
>
> wouldn't it be better to increase the minimal size of swap?

why not remove swap from the standard list of partitions?  it is
a very rare machine today that needs swap, and due to the locking
in port/page.c it's pretty easy to wedge your machine with it.

the oldest bootable machine i have is a pentium ii (dual processor)
with 256mb of memory.  even that needs no swap.  last time i was
at the computer history museum, i saw one of its processors on display!

i'm sure there are some older machines with less memory still running,
but right now we're just making the common case hard by configuring
swap.

- erik

Re: [9fans] Fossil defaults (again)

[tlaronde]

On Sun, Jan 08, 2012 at 10:08:54AM -0500, erik quanstrom wrote:
>
> this only applies to automatic partitioning.  if you set up
> partitions manually, it doesn't apply.  what do you think
> automatic partitioning should do?  should it only be of use
> for installing the distribution, or should it be of general use?

Then---when proposing a new set of installation scripts---I will modify
the scripts to catch the size problem.

>
> why not remove swap from the standard list of partitions?  it is
> a very rare machine today that needs swap, and due to the locking
> in port/page.c it's pretty easy to wedge your machine with it.
>

Since the installation is for the general case, probably swap can be
dropped.

[BTW, the scripts in sys/lib/dist/pc/inst/, for newcomers, are IMHO
opinion a good place to start to learn more about Plan9.]

--
        Thierry Laronde <tlaronde +AT+ polynum +dot+ com>
                      http://www.kergis.com/
Key fingerprint = 0FF7 E906 FBAF FE95 FD89  250D 52B1 AE95 6006 F40C