Trying to identify SCSI cables

"Jesper Monsted" wrote in message 4.163
"Folkert Rienstra" wrote in :
They'll work with any (high voltage) differential device.

Yes, can't possibly work with LVD devices, can it. The cable would know.

Since LVD doesn't allow cable lengths of 19 meters,

Actually it allows 25 meters, point to point.

no, it wouldn't.

So, yes it would.

On Wed, 10 Dec 2003 00:24:19 +0100, "Folkert Rienstra"
wrote:


Oh, dear. Oh dear, oh dear, oh dear.

Do you KNOW what HVD means?

Yes Malcolm, you know I do.

No, I don't. And you aren't demonstrating any evidence of so knowing
now...

Before you go around calling people "clueless", perhaps, Folkert
Rienstra, you should have a clue yourself.

I have a 4 year existence in this newsgroup that says that I do, Malcolm.


Ooh! 4 years *in this newsgroup* (which newsgroup, by the way, since
there are two in the list)...

Me, I've been posting in comp.arch.storage for nearly 10 years....

BUT, and much more significantly, I've been making and selling SCSI
devices, including HVD ones, for at least that long.

The particular clue you should have is in section 5.1 of X3.131-1994,
which is the designation given to revision 10L of draft X3T9 ANSI
standard better known as SCSI-2 when it was approved by the American
National Standards Institute's Board of Standards Review on January
31, 1994.

Isn't that that experimental standard that was never finished and was
overtaken by the SCSI-3 standard, Malcolm?

No, that's ANSI X3.131-1994, the SCSI-2 standard.

Which was approved, and, gosh, what a surprise, I've even given you
the *date* that it was approved.

Now, faced with your evident ignorance, I'd perhaps add that *all* of
these standards start off as drafts, and at some stage a draft becomes
final. The draft that became final happened to be called rev. 10L.

I.e. the damn spec, Folkert Rienstra.

A draft, actually. An experimental and never finished spec, Malcolm.

Nope. The ANSI standard, approved and adopted.

It's called SCSI-2.

It was approved on Jan 31, 1994...

For your edification, here's that section:

[ Begin ]
5.1 Physical description
SCSI devices are daisy-chained together using a common 50-conductor A
cable and,

optionally, a 68-conductor B cable.

Which shows the experimental stage of that spec, doesn't it, Malcolm.

Both ends of each cable are terminated. All signals are common between
all SCSI devices on the A cable. In systems that employ the wide SCSI
option, wide SCSI devices additionally connect to the B cable. Various
width SCSI devices may be mixed.

For your edification, Malcolm, that means both A and B cables connect
to a wide device. You ever seen one of those, have you?

Yup. 8 bit A cable, 24 bit B cable.

Made by my current employer, as a matter of fact.

It is the only *very wide* device I've worked with. But it does work,
using a VME-based HBA from Ciprico, IIRC.

NOTE 1 An alternate 16-bit single-cable solution and an alternate
32-bit solution is being defined and the B cable definition will be
removed in a future version of SCSI.

Which in practice happened much sooner and may or may not have
found it's way in the final SCSI-2 spec that isn't available for free.

It didn't. As a subscriber to the standard, I can assure you of that.
The as noted above, draft revision 10L became the standard...

You appear, now, to admit that you don't have a copy of the standard.

Which makes your ignorance more understandable, but not any more
forgivable.

Two driver/receiver alternatives are specified:

a) Single-ended drivers and receivers, which allow a maximum cable
length of 6 m (primarily for connection within an enclosure).

Not for Fast SCSI, that is (Fast-10).

Yes, for Fast SCSI, with a 100ns synchronous period .

b) Differential drivers and receivers, which allow a maximum cable
length of 25 m.
The single-ended and differential alternatives are mutually exclusive
on the same physical bus.

NOTE 2 Use of single-ended drivers and receivers with the fast
synchronous data transfer option is not recommended.

Well, that again really shows how much behind (or incomplete) that draft
was.

It's not the draft, it's the standard!

Not only did Fast SCSI use SE drivers, Ultra SCSI (SCSI-3) after
that did use SE drivers. What changed was the use of active negation
drivers in SCSI-3.

That is true. It is *also* true that the use of SE drivers *was* NOT
recommended by ANSI.

Which is why we now have LVD, see?

The 6 meter limit is for Asynchronous SCSI (Fast-5).

Nope.

Go check the spec. Here's a hint: the bit you want is in section 5.

Further, the idea that async was *ever* referred to as "Fast-5" is
pure nonsense: see Section 5.8 of the spec, which states

"When devices negotiate a synchronous data transfer period of less
than 200 ns they are said to be using fast synchronous data
transfers."

And mo asynchronous transfers don't *have* a specific rate
associated with them, *because* they are asynch.

The whole point of 10Mxfer transfers is that the transmitting LU can
clock data out at a fixed, negotiated clock rate (up to 100ns for
SCSI-2). Async has no associated clock: the data is latched by the
REQ and ACK lines. Thus the maximum data rate of Async SCSI is
determined by the minimum deskew times. If you do the math, and
assume that each device is capable of responding instantaneously to a
change of signal, you find that the absolute limit defined by the spec
is one deskew delay plus one cable skew delay, for a total of 55ns per
transfer. Which is rather better than minimum 100ns per transfer
provided by Fast SCSI..

Of course, to be safe, wise implementors will add some more delays,
but they aren't required to..

Which leads to the fact that SCSI-2 is limited to 10MXfers/sec for
synchronous, and SCSI-1 and SCSI-2 are both limited to 18MXfers/sec
asynchronously...

[End]

That paragraph "b" is the bit that should be causing you embarrassment
right about now, Folkert.

Not any more than it should do you on paragraph a), Malcolm.

Nope. It's the spec, Folkert. The fact that you've obviously never
*read* the thing is your problem.

Me, well, I've designed and debugged devices for a living (and kinda
still do...)

By the way, note that the people who wrote the damn spec recommend
*against* using single-ended Fast SCSI, which makes your initial
mis-statement even more ironic!

Or your's moronic, Malcolm.
Notice that "Fast SCSI", Malcolm? As opposed to Asynchronous SCSI?

Nope. Sync as opposed to Async, Fast as opposed to... not fast.

The spec, which you admit you've not seen, makes this quite clear.

Please *try* to learn *something* before you criticize those of us who
do this for a living!

And even if you do, you'll see that the single-ended max length is 6m,
or about 18ft, not the "10ft" that you claim.

That's for async, Malcolm, not synchronous Fast-10, better known as Fast SCSI.

Get the spec. Read it. Apologize at your leisure.

Go see the SCSI-3 drafts, which build on the SCSI-2 spec and should there-
for be closer to the final SCSI-2 spec than that SCSI-2 draft ever came.

Don't be silly! Go *learn* something...

You do of course know where to find the SCSI-3 drafts, don't you, Malcolm?

Yup. Middle shelf of the bookcase next to my desk. They're next to
the SCSI-2 Standard.

(I say "they" because, unlike the monolithic SCSI-2 standard, SCSI-3
is actually a family of standards; I'm particularly interested in the
SAM, SSC, SMC, SPC, SPI, and FCP).

Feel free to apologize any time you like!

Likewise, Malcolm.

I have made no factual errors in this thread. You have. You have
further claimed I was incorrect, even though you admit you haven't
read the spec...

My statement may have been overgeneralized, but so was yours.

No, mine was accurate.

But I'll let you into a bit of a secret: when I was working for EMC
(circa 1995) we (being the open storage engineering group) tested, and
approved, SCSI cables about 50m long. Marketing eventually rejected
them since they cost us about 20 times as much as our regular cables,
and felt that they didn't provide enough benefit to be worth the
trouble.

But another bit of a secret: the Adaptec 2944 PCI adapter can drive
20m SCSI cables quite happily, as can various Sun PCI cards. The
Qlogic 1240D has problems with them.

Worse is that you don't seem to have grasped what SCSI is all about.

Tell you what: why don't *you* read the spec, and then come back to
me.

And while you're at it, develop and debug a SCSI HBA. I'm personally
partial to the LSI Logic family of protocol controllers, but feel free
to use whatever you like.

Malc.

On Wed, 10 Dec 2003 00:25:12 +0100, "Folkert Rienstra"
wrote:


"Jesper Monsted" wrote in message 4.163
"Folkert Rienstra" wrote in :
They'll work with any (high voltage) differential device.

Yes, can't possibly work with LVD devices, can it. The cable would know.

Since LVD doesn't allow cable lengths of 19 meters,

Actually it allows 25 meters, point to point.

Except that it doesn't. SIP-5 states that for Fast-320 the limit is
20m...

no, it wouldn't.

So, yes it would.

It might. Not would.

The issue is the electrical properties of the cable. HVD cables (such
as this one) had far fewer requirements than LVD has. If this cable
happens to meet those requirements, you should be fine (since a 19m
cable even fits within the Fast-320 SIP-5 limit of 20m). If it
doesn't, you won't be fine.

Malc.

I have a 4 year existence in this newsgroup that says that I do, Malcolm.

A "google groups" (formerly DejaNews) search finds one of my postings:
From: J. Jonas )
Newsgroups: comp.periphs.scsi
Date: 1992-11-11 04:14:36 PST

you don't have seniority.

I bought some long long SCSI cables on ebay really cheap
since they're just hard to use.
I'll cut them shorter as needed.

"Jeff Jonas" wrote in message
I have a 4 year existence in this newsgroup that says that I do, Malcolm.

So Jeff, any particular reason why you snipped that out of it's context?


A "google groups" (formerly DejaNews) search finds one of my postings:
From: J. Jonas )
Newsgroups: comp.periphs.scsi
Date: 1992-11-11 04:14:36 PST

you don't have seniority.

I didn't claim seniority but it sounds to me that you now *do*.
Perhaps I'm better off not having your seniority.


I bought some long long SCSI cables on ebay really cheap
since they're just hard to use.
I'll cut them shorter as needed.

On Wed, 17 Dec 2003 00:21:28 +0100, "Folkert Rienstra"
wrote:


"Jeff Jonas" wrote in message
I have a 4 year existence in this newsgroup that says that I do, Malcolm.

So Jeff, any particular reason why you snipped that out of it's context?

The context being, of course, that despite having loitered for the
past 4 years, Folkert is aggressively ignorant!

A "google groups" (formerly DejaNews) search finds one of my postings:
From: J. Jonas )
Newsgroups: comp.periphs.scsi
Date: 1992-11-11 04:14:36 PST

you don't have seniority.

I didn't claim seniority but it sounds to me that you now *do*.
Perhaps I'm better off not having your seniority.

You may be better off refraining from stupid claims.

And looking at specs.

Malc.