[Rule-list] The state of the installer message

[Devon]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


OK, a few random thoughts on progress and direction.

I think we all agree that it would be wonderful if Red Hat were to 
include the ability to install in lower memory machines. For this to 
happen, I'd think any changes we make to anaconda would need 
to isolated, perhaps as a new install method.

We can easily modify the MIN_RAM requirement, for example. Red Hat, 
however, isn't going to want to see that lowered overall. They don't want 
people trying to do a full install in 12M of RAM (and they will), because 
when it fails, (and it will) it would reflect badly on Red Hat. Stock 
holders don't like bad publicity. ;)

To solve this, perhaps we could add a little code to the memory test. If 
memory < 10000, sorry, can't be done, exit. If memory < 17000, display 
disclamer, offer only the 'lowmem' install classes, and a slimmed down 
comps file. Offer to quit or continue. 
Just something to consider.

Another possibility that shouldn't be too difficult to add is the ability 
to skip partitioning, and perhaps formatting. (This is already done on a 
hard drive based install, as you can't format the drive containing the 
ISOs.) This would speed things up in low memory situations, and allow 
swap to be turned on very early. Perhaps if usable partitions exist, 
offer to format them, offer to turn on swap, etc. I haven't really 
thought this through, but it seems like a path worth consideration. The 
ability to create and format these partitions outside the installer would 
be useful, of course.

The reason I bring this up is that I believe we have hit the low end of 
the memory limit, while continuing to use anaconda and a stock Red Hat 
distribution. I believe that limit is going to be ~12M. Yes, with more 
changes, we could possibly slim it down more, but I doubt Red Hat would 
be interested in supporting that ability. On top of that, how many end 
users would be willing to endure a basic install that takes 12 -16 or 
more hours, resulting in a machine that uses 98% of the cpu time just 
managing swap? Swap is a wonderful thing, but it isn't a substitute for 
RAM. Can you imagine trying to compile a kernel or rebuild a large RPM 
under those circumstances?

I'm not trying to be negative, just realistic here. Red Hat has to 
support what they sell, and isn't likely to want to deal with the support 
calls a low memory install might generate.

Thoughts, opinions, counter claims, flames?

- -D

- -- 

pgp key:  http://www.tuxfan.homeip.net:8080/pgpkey.txt

- --
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iD8DBQE8evH2eMAUbzJhSVcRAhjqAKCFiu9JBemsksFtcQ9qPU71QRE26QCguyw6
SX5wyxQgH6TAoKRHpAz9F0s=
=6Eys
-----END PGP SIGNATURE-----

_______________________________________________
Rule-list mailing list
Rule-list at mail.freesoftware.fsf.org
http://mail.freesoftware.fsf.org/mailman/listinfo/rule-list