10mm or thicker float is the thing to go for if you are buying it, if you can rustle up a freebie all the better. Our customers have had various reactions from glaziers, from 'no problem we will put an offcut aside for you' to quotes of up to £40. - I know who I'd be going back to when it comes time to get the house re-glazed!
Grades wise I have settled on 40, 30 and 5 micron for general sharpening, the next most common is 30, 15, 5 and 1 set out lengthwise on the glass, I believe this is the option favoured by the Barnsley Workshops but scary sharp is a very flexible system so you can have a play and then adapt it for your own needs. I'm the first to admit that 0.3 micron is overkill but it does remove metal and will squeeze the last ounce of performance out of a blade, if you have a really critical cut to make then it's nice to have a sheet or two on hand.
For grinding I use 100 micron stuck to MDF; plenty flat enough for grinding and you can bin it when you are finished, the 100 micron sheets are not available in PSA so I stick them down with PVA or a smear of epoxy. The 100 micron removes material rapidly but in a controlled way, it will produce a beard of iron filings on the blade and get it hot enough that you can't hold it in about a minute, which is a useful safety valve for not overheating the steel. If you are just dressing the primary bevel you will probably be done long before that happens anyway.
Hollow grinding simply means a surface that is intentionally made concave. In chisels you can have hollow ground bevels, as Alf describes above, which are often preferred by freehand honers as it gives them two lines to register against the surface of the stone (that is if they are not using the rounded bevel method as advocated by Jacob although I'm guessing you could combine the two quite happily if you got a really bad ding in the edge).
Hollow ground backs are split into two distinct types but the principle is the same, the arrangement is self-jigging and greatly reduce the amount of effort required to produce an accurate surface:
First there is the Japanese Ura where a shallow oval is ground into the centre of the blade. This gradually recedes ahead of the cutting edge each time you polish off the burr so it will become circular and then oval the other way as the chisel wears but never actually break through the edge.
Second is the English hollow ground chisel (i.e. Ashley Iles) where the non bevelled side is concave in its length by a couple of thou, this is not enough to make the chisel dive in the cut by any amount that is significant in even exhibition quality work, but it is just enough that the underside of the bevel can be polished in isolation from the rest of the blade. Barry Iles and Tony the foreman spent about 18 months perfecting the hand grinding technique for achieving this reliably and ended up achieving half the tolerance that I had asked them for.
It makes them very fast to hone and because you are working such a small area you can eliminate the scratch pattern from the previous grade easily and completely, you can typically get even a wide blade from box to bench in under 3 minutes. If the shiny bit starts to get too big, working the middle of the blade at an angle across the edge of a sheet of 100 micron or across the width of a coarse stone (David C's method) will reinstate the very fine curvature.
The only other chisels I have found with a consistent hollow in are Narex, I believe this is caused by shrinkage during heat treatment - if you've ever cast lead in the frog of a brick you will see how the billet shrinks back in the centre. This is usually just a fraction deeper than the AI's and in an oval shape like the Japanese chisels, it is not engineered to be there, just a happy by-product of their accurate grinding and austempering hardening process.