The guideline fro kernel 2.0 was twice RAM up to 127M (biggest swap available.
The guideline for kernel 2.2 is 2xRAM up to about 250M, thenequal to or even smaller than RAM is probably OK for desktops (might be better to have more with some memory-hungry servers). For kernel 2.4 up to 2.4.13 or so, the volume manager used swap lots of swap and even more swap. Give swap all you can spare. Supposedly the vm doesn't hit the disk as much since 2.4.13, but lots of swap is still good. The questions arise One Swap or more? Where to put swap on the disk? IBM has a very very nice set of whitepapers on SWAP, and it is good reading for the moderately computer literate to whom I recommend it. Just search their site for 'linux swap'. Basically put SWWAP as early on the disk as you can to shorten stepping time and use at least one SWAP partition on each drive. You can prioritize which are used first, but if you leave the priorities equal something amazing happens--the SWAPs use striping just like a RAID0 You might arrange priorities for each swap if you had say three of them and one was on an old disk drive which was slower than the others (lowering that SWAP's priority takes it out of striping and lets it be used only when others are full, which means the striping performs really fast. Civileme
Want to buy your Pack or Services from MandrakeSoft? Go to http://www.mandrakestore.com