This is incorrect. You want your final DVD to be a 24fps progressive source. This is what the actual film is, and what a retail DVD would be. It's also fairly simple to achieve with AVIsynth. The Laserdisc is 30fps interlace, which means 1/5th of the data is duplicates of other fields, and an interlace source will not compress well anyway.

Using AVIsynth, you deinterlace to give you 60fps, with 2/5ths of these being unique images. These are the real movie. Taking them, you have your 24fps film. If done carefully and properly, it will produce images practically identical (no blending or sync issues at all) to the original film before it was altered to make the laserdisc. You end up with a better image that will also compress better (preserving more image quality), and doesn't rely on hardware to deinterlace it (which can vary in quality).

You really want a non-interlaced image for running through filters and MPEG encoding - it'll produce the best quality, and best compression.

Of course it depends on the -original- source. In this case we're talking about a 24fps original source (film)... we don't care what format the laserdisc is in, except to undo the damage they did when making it.

Yeah, sorry Doctor M, but any project worth its salt does an IVTC process first to completely reconstruct the original 24fps film frames.
So farsight did this because he *did* know what he was doing

When the guides talk about 'The source' being interlaced, they mean the original source. In our case the original source is film, whic is 24 progressive frames per second, which was then transferred to laserdisc and made into a 30fps interlaced mishmash.
The IVTC process recreates the original film frames with no quality loss by extracting the matching interlaced halves and putting them back together.

If the 'original source' was video (i.e. it was shot using an interlaced video camera in the first place) then you cannot do an IVTC, and any form of de-interlacing does cause a quality degredation as there are no original progressive frames to get back to.

So for stuff that originated on film, doing an IVTC is a must, for stuff shot on video, you can only get a compromised image by deinterlacing and averaging and so forth.

Actually I found the IVTC built into the later versions of virtualdub never has a problem. I think it is nice and dumb and just removes based on a set pattern and doesn't try to 'guess' like the various 'automatic' avisynth filters do.
It only gets out of whack when there is an error in the source footage, in which case you just start it again from the problem frame.

Using AVIsynth's DoubleWeave, followed by Pulldown, worked perfectly for me. You just have to manually figure out which field comes first (since you can't rely on the LD to not have errors), and must also figure out which 2 of every 5 frames are unique (easiest to do in sections with fast motion, since 2 frames will be interlaced, and 2 others will be identical).

Bottom line - it's a pain.

Okay, this is nothing to do with Farsight's set, but it kind of follows on from the above: how does on-the-fly PAL to NTSC conversion work? Are PAL DVDs generally encoded at 24fps progressive, with the PAL player converting to 25fps? If so, does the PAL/NTSC player apply 3:2 pulldown to get to 29.97 at true speed, or is it somehow still played at PAL's 104% speed?

All decent film DVDs are stored as 24fps on the disc itself. No DVDs store anything as progressive, it is an interlaced system - but the two fields stored make up exactly one film frame, so the end result is the same really.