13 Aug
2012
13 Aug
'12
10:34 p.m.
On Sun, 12 Aug 2012, Dave McGuire wrote:
On 08/12/2012 08:13 PM, Mouse wrote:
A godsend...to hardware makers and others who benefit from forced
obsolescence.
I'm talking about its technical implementation, not the business
implications. ATA was very much at the "end of its road" in terms of
its design by, say, 1993 or so. We saw how long it lasted after that,
for no good reason.
Meh... You could say the same thing about single ended SCSI though. The
reason SCSI was able to last is that it continued to evolve. Parallel ATA
on the other hand, evolved differently, mainly because of the consumer
computer market.
The irony is towards the end of ATA's widespread use, it had evolved to
the point where it had a lot in common with single ended SCSI, but because
of the way of had been forced to conform to existing standards, it
retained some major flaws.
Look at the UDMA 80-wire cables for UDMA-66 and faster. The industry
didn't want to change from the 40-pin 2.54mm connectors, but needed a
ground wire between each of the data lines to avoid crosstalk. SCSI on the
other hand had done this for years, even with good old 50-wire ribbon
cable. In order for ATA to do this so late in the game, special wire and
special connectors were required, which were not only much more expensive
and difficult to assemble, but they had more long term reliability issues
as well.
ATA drives were also speaking SCSI towards the end, just look at all the
widespread implementation of ATAPI in not only cdrom and dvdrom drives,
but even hard disks.
At least people wised up and decided maybe the primary ATA/IDE hard drive
really should be on the far end of the ribbon cable. With 40-wire cables,
the convention had been to put the primary drive in the middle (which can
be problematic since you can get signal reflections in that loose ribbon
cable "tail"). When the industry switched to the UDMA 80-wire cables, the
convention changed to putting the primary drive on the end of the cable
instead of the middle (and this was enforced with the use of those
proprietary ribbon cable connectors since the primary and secondary
connectors omit different pins).
> [...run eight parallel data pairs...]
The problem is that in general you can't [as speeds rise, skew and
crosstalk get bad]. That's why we have gone SATA rather than PATA.
Well that
was one of the reasons. ;) ATA/IDE was one of the
worst-designed interconnects in the history of this industry. Even
the PC weeniez knew it had to go. SATA, for all of its faults, is a
godsend!