I'd like to use SATA for three reasons.
Firstly, it's the easiest way to get everything on its own channel as a master (there's no slaves on SATA). That means no multi-tasking issues. Even though four of the seven drives in my current setup are SCSI, I'm forced to put the two DVD drives on a single IDE channel.
Secondly, you don't have the cable length issues on SATA that you do on PATA - PATA is limited by specification to 18 inches (45 cm) - though 24 inches (60 cm) cables are common, they are in breach of the specification.
Thirdly, there don't seem to be any PCI-X PATA cards.
You can get little adapters from Highpoint that allow you to use PATA drives with a SATA controller - they're little more than a serialiser/deserialiser chip in a box, so they shouldn't introduce problems (famous last words...!). They cost around US$15 each.
It's not a bandwidth issue - current ATA hard disks don't push significantly at the limits of UDMA-66 unless they're streaming, and even those of the highest areal densities are currently within the limits of UDMA-100 (the same is just as true of drives like my expensive Seagate Cheetah 15K.3 hard disks, 15000rpm U320 SCSI - my two Cheetahs are capable of an internal speed of up to 86 MBytes/second, but outside certain applications like video, drives are called upon to seek a lot - seek time, of course, is where drives like the 15krpm Cheetahs excel). Current optical drives are fine on UDMA-33 - the same will be true of 16x DVD writers. Some support UDMA-66, though there's no real reason to do so, especially as IDE cannot multi-task two devices on the same channel.
I've put my current drive setup at the bottom of this message - if it wasn't that I was using SCSI, I think SATA would be the only way to not have a complete mess.
So far as dual layer writers go, they're going to be slower than writing two single layer blanks at launch, and media availability seems destined to be poor for several months.
If you are wanting a new drive now, then if you can hang on until the first dual layer products are available that probably makes sense - it seems that DVD+R9 capability is going to add little to the price, and if you buy wisely you'll have a drive that's just as good if not better as the current leading edge single layer DVD writers.
There could well be the kind of compatibility issues you fear at first - media not being readable in all drives, and certain drives only writing certain brands of media. Add that to the limited availability of media to start with, and the possibility that we might have 4x DVD+R9 and even DVD-R9 by the end of the year, and I'm going to sit still with my ND-2500A. I bought when I needed DVD capabilities earlier this year (I'm now handling files that are just too large for CD-R), recognising that I was buying into a pretty mature technology that would do all I needed at a pretty decent speed for less than GBP100 for the drive (I paid GBP88).
If an official ND-2500A to ND-2510A upgrade is made available, I'll probably buy it if it's not too expensive, but more to get the most out of my existing investment than for any other reason (I doubt it will be free unless there's no extra licence fee from the DVD+RW Alliance for DVD+R9). I wouldn't be surprised, though, if drives sold as ND-2500A lack the necessary factory calibration data for dual layer writing, defeating even an unofficial upgrade. It's also possible that even if the drive architecture is the same, reliable dual layer operation means that the ND-2510A requires revised optics.
Unofficial upgrades also raise the spectre of worse compatibility problems than early DVD+R9 may have anyway. Turning 5 to 10 expensive DVD+R9 blanks into coasters via a failed upgrade could well be a fair proportion of the cost of a drive with official DVD+R9 support from new.
I don't really need dual layer myself - my main need is for DVD data. I certainly am not interested in unofficial upgrades - I value my data integrity too much.
I'll probably stick with the ND-2500A for a year or so before considering a new optical writer - but I appreciate others have different priorities.
If you're wanting a new DVD writer, then the first generation dual layer products are likely to be just as good if not better as the current single layer writers, for not much more money. However, for those who are intending to write a lot of dual layer blanks, I wouldn't be surprised if in another 6-12 months, they'll be wanting to upgrade again. 2.4x to 4x is a big drop in writing time across two layers (I think the first DVD+R9 writers are going to be 2.4x, at least on the second layer if not on both).
I certainly wouldn't want to buy a dual layer drive without seeing the first reviews.
Just as a matter of interest, my current setup is this:
LSILogic 1020 U320 SCSI controller
ID 0: Seagate ST336753LW - 36GB 15000rpm Cheetah 15K.3 U320 SCSI disk drive
ID 1: Seagate ST336753LW - 36GB 15000rpm Cheetah 15K.3 U320 SCSI disk drive
Main swap file, data
ID 4: Seagate / Certance STD2401LW - 20/40GB DDS-4 Scorpion 40 U2W SCSI tape drive
Master: Seagate ST3160023A - 160GB 7200rpm Barracuda 7200.7plus UDMA-100 disk drive
Backups to disk mainly, also scratch space
Master: Lite On LTD163 UDMA-33 DVD-ROM drive
Slave: NEC ND-2500A UDMA-33 DVD+-RW drive
Those six drives are all internal on my Dell Precision 650, using the motherboard controllers (this dual Xeon machine has a SCSI controller on the motherboard).
I added a seventh drive last week:
Adaptec 29160N U160 SCSI controller
ID 6: Plextor PX-W1210TS - 12x10x32x Ultra SCSI CD-RW drive
I had this Plextor drive in my old computer, and the Adaptec PCI SCSI controller was a spare I had anyway. I managed to get an an external SCSI enclosure brand new last week from eBay for GBP10.
I have a load of audio CDs (at least 30, possibly more) to write in the next couple of weeks on behalf of the composer/performer, and I decided I'd rather write them on the Plextor than use my ND-2500A. I've used the Plextor for audio CDs with good results before; if I chose to write the discs on the ND-2500A I wouldn't want to use more than 16x CAV, so the 12x limit of this drive is no disaster.
If I wasn't using SCSI, SATA would provide a much neater solution for that number of drives on one machine!