Once upon a time, I loved SCSI. (Was: Help! Is this kernel or hardwareproblem?)

[Benjamin Scott]

On Nov 27 at 1:59pm, Ken D'Ambrosio wrote:
> You took your AHA-1542B card, plugged in your stuff, and It Just Worked.

   Heh.  Sure, if you stayed in the little pee cee play-pen.  Throw a 
Micro-VAX or an SGI into the mix, or try getting one of Apple's goofy designs 
to work with anything else.  :-)

> There were, hell, no more than two or three different types of SCSI back 
> then ...

   Erk.  See below.  :)

> ... SCSI-1 ...

   Do you mean the SCSI-1 50-pin "Centronics" connector, or the SCSI-1 50-pin 
IDC "ribbon" connector, or ...

> ... the funky Exabyte Andataco?) four-quadrillion-pin-highly-proprietary 
> thang ...

   I'm guessing you mean the so-called "DB-50" connector, there.

> ... and Mac's 25-pin-looks-like-a-serial-connector bit).

   Apple Kludge.  It was "SCSI" in the same way that "Windows 95" was a 
multi-tasking, memory-protected operating system.

> Then it just got weird.  You had SCSI-1, SCSI-2, SCSI-3, wide, LVD,
> active, passive, blah, blah, blah.

   Most of the confusion is perpetrated by marketing weenies, combined with 
some standards documents that failed to put forward useful terminology, and 
the fact that SCSI has evolved incredibly over the years.

   One thing you have to realize early on: SCSI-1 and SCSI-2 both refer to 
*revisions of the SCSI standard*.  They are not connectors, or speeds, or any 
other damn thing the vendor catalog says.  SCSI-3 (or -4, etc.) don't even 
exist in any official capacity, despite the fact that catalogs love to use 
them as if they did.  Anything that talks about a "SCSI-1 connector" or 
"SCSI-2 speeds" or "SCSI-3 anything" is using non-standard terminology which 
is very inconsistent applied across the industry.

   (Side rant: People often frown when geeks like me make a big deal out of 
using proper terminology.  Well, this confusion is a perfect example of what 
happens when people don't use proper terminology: Nobody can freaking talk to 
anyone about anything, because everyone's terms mean something different! 
Blech!  Terminology is *fundamental* to communication.)

   Anyway, if you want to know, there are a couple of quick reference 
documents I've been maintaining for years now.  I just uploaded them to some 
personal website I have.  Feel free to snag copies (I don't guarantee those 
files will remain available at these URLs in the future).

http://bscott192.home.comcast.net/scsi/SCSI_Parameters.html

http://bscott192.home.comcast.net/scsi/SCSI_Connectors.html

   Since it isn't covered in those documents yet: The SCSI signaling on a 
SCSI bus is generally either SE (Single Ended, the original) or LVD (Low 
Voltage Differential).  LVD is "better".  Anything that supports LVD should 
be able to fall back to SE automagically if needed.  This does mean that one 
SE device on a bus can slow down everything.

   Oh, and again, see http://www.scsifaq.org for all you need to know.  :)

>  And ALL the damn cables cost $100/ft.

   Well, if you want to able to attach 16 devices to a bus that's 40 feet long 
and transfers 320 megabytes/second and running your mission-critical data 
system, then, yah, the cables are going to cost a bit more then the 
cans-and-string one finds at the local PC computer show.  ;-)

> And terminators!  Oh, I love terminators.  Otherwise known as "the things 
> that stick out the back of your machine to get snapped off when you try 
> moving stuff". [Ed.'s note: dashes changed to spaces-and-quotes]

   Interesting.  I never really had that kind of trouble with them.  In my 
experience, the think cables get in the way (and get pulled off) before the 
terminators do.

> And then came SATA.  Or, more specifically, SATA II.  SATA II has done
> some nifty stuff.  It's cherry-picked some of the best features of SCSI,
> while leveraging (damn verbing...) IDE's dominance on the desktop.

   SATA is definitely attractive, although I suspect as time goes on, many of 
the issues that plague SCSI will effect SATA.  I've already seem grumblings 
about sub-standard cables, devices which do not implement the full feature 
set properly, confusion over external connections, and so on.

   TANSTAAFL.

   FYI, the SATA standard adopts several aspects from the modern SCSI 
standards documents.  It can be argued that modern SATA has more in common 
with SCSI then it does with the original parallel ATA standard.

> I still dislike the connectors -- they just don't *click* into place like I 
> imagine they should ...

   I've read that they are (or have) creating a new connector type with a more 
positive locking mechanism, to help keep the cables from being pulled off 
accentually.  Of course, that's another connector variant to add to the 
original internal variant, the eSATA official external variant, and the 
various non-standard external variants introduced by vendors.

   Hmmmm, this sounds familiar.  :)

> -- but I have to say, routing SATA to 16 drives in a cramped chassis beats 
> either IDE or SCSI hands-down.

   I'll take a single SCSI parallel cable connected to a SCSI SCA backplane 
over 15 individual SATA cables any day.  ;-)

   Seriously, I like SATA.  It's about damn time the commodity IBM-PC started 
to get some of the features everybody else has had since the 1980s. 
IDE/ATA/PATA/whatever should have been fixed a long time ago.  And since Dell 
and HP and Gateway and so on are shipping a bajillion of the things every 
year, we get economies of scale we never got with SCSI Parallel drives. 
It's all to the good.

   I just don't think we'll ever do away with the problem of 
(sub|non)-standard and confus(ing|ed) vendors.

-- Ben "SCSI isn't scuzzy" Scott