DVD-A vs. SACD

I'm afraid of getting a bunch of "We've already talked about this" rotten tomatoes thrown at me, but there isn't a search function here and I hate scrolling back through old pages.

I have read two slightly contradictory articles on the issue. The first one, here http://www.audiorevolution.com/equip/dvdavssacd/

says SACD is better but only under the circumstance that no DSP is done to the material after recording, which means almost never.
To fill in what he says with a few more facts. Originally DVD-A was going to be 3 channels 24/96 and 3 channels 16/48 in surround mode and 2 channels 24/192 in stereo. The sampling rate of stereo was set so high to keep the bit rate constant. Meridian Lossless Compression came in and made it so all 5 of the main channels could be recorded 24/96. He is right that 24/192 does not increase perceptual sound quality.

The other article
http://sound.westhost.com/cd-sacd-dvda.htm
is scathing towards SACD. But there are a few questions I have.
One, he talks as if the only kind of DACs on the market are bitstream and "multibit." If you look at Burr-brown's product sheets you will see 3 types: DSD (bitstream), delta-sigma (the most common type on middle and high end consumer products), and "advanced segment" which are right in the price range he talks about for "true multibit" DACs.
Similarly, I once read a passing comment on the laserdisc newsgroup that a tech was disappointed when he opened up a MLD-7020 and saw that McIntosh did not replace Pioneer's "bitstream" DACs with true audiophile DACs. Elsewhere I have learned that delta-sigma DACs typically have an internal resolution of 4 bits (from someone who had not studied conumer audio products in many years, however).

Thinking in terms of my own theory, if you take 24/96 and change it into 1 bit oversampled 24 times, I'm not quite sure why that wouldn't work. I see 2 problems. Problem 1 is that SACD is oversampled 29 and 5/30 times. I don't know a theoretical model where that gets you anywhere.

And, I can see how reconstructing a DSD waveform would produce more of a triangle wave rather than a sine wave, since you'd have a bunch of extra data points between the critical points. But I really don't know squat about the hardware implementation of DAC design, so I don't know if it really works like that. And my vision of filling in the gaps with extra data points only works for integer multiple oversampling, so there are some things I am not seeing.