How do you scan for web-publishing?

Hi, I have been scanning different size of filmsheet and prints lately and am experiencing some problems when I want to use the result for web-publishing.
I scan 6x6, 4x5" filmsheets and 4x5" contact prints at 900-1200 dpi to get a reasonably large tiff-file of say 50-90 Mbytes. I then import the file to Photoshop and resize it to be 600-900 pixels wide with rather good results.

When I scan an 8x10" print, I scan at only 300-600 dpi to get the same tiff-filesize, but when I resize to 600-900 pixels the picture seem to loose much more of sharpness than the 4x5" scan and also when I compare to the print itself. (It seems that the resize algorithm chooses randomly which pixels to keep, which in turn makes straight lines look blurry etc)
This is so disappointing that I hardly want to scan any enlarged prints and instead produce dedicated contacts to scan...sometimes I also dodge/burn the contact, and I do not like doing this artificially in PhotoShop.

Should I scan the 8x10 print at lower resolution to fit the websize from the start or do I need to step up further in dpi or...
I would appreciate some help in this topic. Thanks, Patrik.

FWIW, when I reduce larger scans for Web presentation, I do so in steps of about 50% or less, and apply a light touch of unsharp masking at each step.

I'm always working on maximum manageable size... for me that is 200-300 MB file (could, and usually is, raise up to 5-6...7 hundreds of MB with layers). When I downsample image I do similar as Raplph -- in smaller steps. But I usually unsharp different portions of picture at different stages... for example, If I have some trees and clouds, I will sharpen trees more often than clouds (taking care not to oversharp!). but if clouds have more "texture", I will take give them more details. usually, I give on first few steps more radius in USM and less percentage (say 30-40% and 1.5-3 pixels radius) and much more percentage on smaller radius with fine tuning threshold when I came to smaller size of picture (120-180% and 0.3-.0.5 radius wit 3-4 levels for threshold). Sure, all that depends of(on?) particular scan... Hope that helps a little.

Thanks Ralph and Janko, I will definitely try your workflow...I used to only adjust color and Brightness/contrast in the original scan, then resize only ones to the target size and then USM for sharpness. It sounds better to do as you propose. (I do remember reading a PhotoShop tutorial saying that the first thing to do was to resize to target size and work from there...strange?).

As for USM, I find it hard to know when to use which settings, i.e. %, level etc. what works in one size does not have any effect in another.
So again, thanks alot, Patrik

The PS book or tutorial includes a formula for calculating optimal settings. I use 45% and 0.4, which works well at an image size of 600 pixels (long side), and will occasionally apply it more than once at larger image sizes. The "last filter" memory slot stores that, so it is convenient when doing multiple steps, avoiding having to deal with the unsharp mask pop-up each time.