ImgBurn 2.5.8.0  Win 9x/Me/NT4/2k/XP/2003/Vista/2008/7/8/10 (plus all 64-bit versions). If you use Wine, it should also run on Linux and other x86-based Unixes
   NOTE: Use the 4.71.2 ASPI layer, NOT SPTI

Windows NT 4.0 and Optical Media

"Some software may attempt to access optical drives using SPTI (SCSI Pass-Through Interface). This programming interface was created by Microsoft to allow direct access to hardware like optical drives. Unfortunately, the implementation in NT 4.0 is very poor and unstable. Attempting to access hardware through SPTI may result in STOP or Dr. Watson errors. If the option is available, use ASPI access instead."

Burning at 1,800KB/s under NT 4.0

  I had tried using a 3.36 / 4.70 Buslogic BT646S to drive a Plextor PX-W124TSi, but all it would do upon starting to write is fill up the device buffer, and then the device buffer would drain. Once drained, the burn terminated. This may be an issue with the BusLogic ROM pairs.

  I then installed a Corvette with a wide cable, a Plextor PX-W124TSi (term disabled), 50 to 68 pin converter, and left the heatshrinked active terminator at the end of the F/W cable intact. I used the Adaptec 4.71 ASPI package (includes Win2k version) which installs version 4.60 (correct for NT 4.0). I used IMGBurn.  Fired the Plextor PX-W124TSi up, waffled through the IMGburn screens, and waa-laa, we have a successful burn taking place.

Corvette on 95-Y/NT4/Nero 5.10 Success
  This had me going. I attempted to light up a Plextor PX-W124TSi 12/4/32 at 12x off a Corvette. It bombed under Nero 5.10.56 at 12x (burn failed, buffer underrun). I throttled it down to 8x, hung the CD-R and the data drive off the same Corvette, and I was able to burn successfully.

  I did have to install the Adaptec ASPI layer, ForceASPI could not successfully complete the install. Nero tech support, trying to be helpfull, eventually advised me to enable DMA on the drives, like I was using W9x/XP or other kiddie OS. Hint: there is NO setting under NT to enable DMA. If you run NT Diagnostics, it reports DMA is not being used by the system. Yet Nero Infotool reports DMA... Go figure.

Burning with Nero 4.0 and WfW 3.11
Ezoto piped up with:
I have Nero 4.0 that works on my 9577BTB.  I use a Ricoh SCSI burner and a 100meg internal zip drive.  I don't have the dos driver for the Ricoh but using the Future Domain SCSI card with the FD drivers in WFW works perfectly.  However after you install it and then run it it asks for the WINASPI.dll file that is in WIN95 (OSR2 Win95_02.cab) but that is easy to fix since I copy it from 95 to the WFW system directory.  After that I successfully burned on my 9577BTB with the 83mhz ODP and 64meg ram.

One Possible CD Burning Configuration
   For multitasking during a CD-burn session on a relatively slow (P90 plus) relatively memory-poor (64MB-128MB) MCA/PCI-bus system, I recommend using a whole separate SCSI adapter (with no non-bus-mastering I/O for the "other" stuff) controlling a drive dedicated to the CD Image and the CD-Writer. 
   With good software that can work out that all operations for the CD-burn are on the same adapter+bus between two devices then the controller *should* be able to do inter-drive operations without anything but tiny CPU and host memory requirements. This is a "special situation" solution that should, in fact, be more commonly known, understood and described in the software blurb for CD-R/RW/DVD-R/RW/RAM packages.
   *Never* take the "minimum requirements" on a package/service's advert to heart. It's based on the poor knowledge and understanding of the technical issues by the "general public". A "fail-safe" cop-out.
-- 
Regards,
        Tim Clarke (a.k.a. WBST)

CD and DVD Musings
I think you need to differ between two points:
a) what you really need from the hardware
b) what the accompanying (Windows-based) software needs

The data stream required for a 4x burning process is about 500Kbps, which can be done with a reasonably configured 486DX33 / 16MB and a sufficiently fast HD subsystem already. But the lot of the fat software packages require Win 9x / ME or even XP - were the latter two are often too big for older machines.

The difference between CD-R and CD-RW burner is mainly just a firmware issue - and a slightly modified laser control. Since CD-RW includes burning CD-R already the manufacturers opted to drop sole CD-R burners in favor of lower-priced CD-RW units. I can see no problem with it. I would always prefer a CD-RW capable burner over one that does only CD-R.

Far better however is a multiformat burner that also does DVD in its various manifestations (DVD-R, DVD+R, DVD-RW, DVD+RW and DVD-RAM ... and Double Layer) ... plus the CD-R / CD-RW stuff. Would be a bit more expensive but more versatile as well. Win98SE is sufficient to run e.g. MyDVD software for creating home-made DVDs My PC300GL 6561 with 333MHz P-II and 128MB does fine here - and it is far from being a "modern" system.

-- 
Very friendly greetings from Peter in Germany
http://members.aol.com/mcapage0/mcaindex.htm

A kicker is a Wide drive and CD burner on a F/W. I had a 20x Plextor (with the HPDB68 port) that refused to run on a Corvette until I disabled wide messages.  Perhaps running the burner and a dedicated HD on the same Buslogic BT-646 would then be the best setup. 

Lack of ASPI Layer
   Known problem. Use ForceASPI to install an ASPI layer on your 9x/NT system. I have used it under NT 4.0 without a problem. This will let you have an Adaptec ASPI layer for a non-Adaptec SCSI adapter. It works with IBM SCSI adapters as well. I know about Spock and the Corvette personally.

The key is the first sector, a kind of "Boot Sector". Not a real one, of course. It is described in the original El Torito (Boot CD) specification, but seems to be forgotten in the CD layout description  and marked as "not necessary". And since El Torito is a specification by Phoenix and IBM, IBM CDs do work - they knew how to do it. The ServerGuide CD has this "Boot Sector". And others, who implemented El Torito based on the specification, just skipped this part.

News!:
   I don't know of WinDOS burning programs like Nero and so on, but mkisofs (from the CDRTools or CDRTools/2 package) generates an almost valid booting table. The problem: It generates it in a file - named boot.cat(alog) by default - on the file system instead of sector 0! We need to patch mkisofs to put the boot catalog into the right position. But first I'll contact the author of mkisofs, maybe he can do it better, faster, and right into the official source tree.

   Writing an utility would be a bit difficult, but not impossible, because it would need to analyze the disk image if it already has this boot catalog. If not, it would need to generate a new one. And if it has, it could copy it to sector 0, and subtract the negative offset to calculate the new relative addresses. But I'm not sure if I can program all these things ... or better: I'm sure I can't program all these things. Maybe someone other wants to try it? *g*

Here are the specs, the boot catalog can be found at section 2: HERE