If you spend a bit of time researching your options for loading OpenStreetMap planet dumps into PostGIS, you soon realise that you have got two competing options: Osmosis or osm2pgsql.

Before you jump in and make your decision, and start churning through 100Gb of data, have a think about why you are loading the data.

If you are loading data to render it using Mapnik to either generate static tiles or as a Slippy map server/service, then you will want to use osm2pgsql.

If you are loading data to have an easily updateable mirror of the OSM raw data structure, then you will want to use osmosis.

Whilst this sounds pretty obvious to anyone who has worked with OSM data, it wasn't obvious to me.  Having spent a good while watching an Osmosis load run, I realised that the schema used is completely inappropriate for standard Mapnik renderings; it stores data as very basic nodes/ways/relations, whilst osm2pgsql will create nice point/line/polygon tables with appropriate tags.

Of course, there is another lesson here in that you should experiment with a small subset of data, and not an entire planet dump!