One method is to dispense with PDF and just view the scanned pages online as
images or OCR'd text or point the user to a directory with the scans
for the document.
He then only needs an image viewer using a lot less of his machines memory.

Large PDF's also cause problems in the viewing computer. I was
reviewing someones
25mb PDF the other day and it peaked at 3.3 gig memory use, which on a 2.5gig
memory box meant it went into swap and slowed to a crawl.
The viewer used there was evince.

I scan to jpg and only produce a PDF if nagged

As I serve from home and the upload is on the slow side individual
pages helps there too.
And when in a good mood I finish off a document thus
where all pages are web viewable. Been too lazy to write a page to
page link on the page
view so far (need a round tuit).

Dave Caroline