[Tfug] Switch problem

[Bexley Hall]

Hi Louis,

On 3/17/2013 12:37 PM, Louis Taber wrote:
> I set up a test bed in my shop with 600 ft of Cat-5.  It
> quite happily auto-negotiated to 100MBit/sec, but would move no data at

OK.  So you came to this conclusion from empirical data.  I
was hoping there was something I hadn't noticed in the written
specification that would have clued you in on this scenario... :<

> all.  Put an old hub at one end to force it to 10MBit/sec and it worked
> fine.  See: http://en.wikipedia.org/wiki/Autonegotiation#Electrical_signals

OK, so in each case, you get a "go" for the "Link" (LED) but the
first case just leaves you with no *legitimate* signal at the far
end.

So, in Erich's case, the machine can think the link "up" and just
never see any "in band" data (e.g., DHCP traffic)  (?)

In which case, forcing the interface to a specific configuration
(my original suggestion) should work?  I.e., don't *let* the switch
"think" the link (end to end) is capable of anything that it *isn't*!

[Though he stated that the slower/older NIC's seemed to be manifesting
the most problems...  :-/ ]

> A side note: I have contended for years that the SOHO market needed to
> handle miss-wired cables and mid-x to reduce returns and customer service
> costs.

I think AutoMDIX adds another set of issues.  E.g., I suspect that was
the problem I was having with my two Gb switches (separated by 8 ft
of cable).  Powering them up *singly* would work but if they both
came up together, they just sat staring at each other...

But I agree with your observation.  A wise manufacturer would design
in a way that would discourage the user from even *suspecting* the
kit (assuming it most usually is NOT the culprit!).  Or, at the very
least, provide a reassuring means by which the user can convince
himself that the kit *is* or is *not* the (part of) problem!

> Fiber will always be wonderful in the right environment.  2 miles off the
> paved road 15 miles from Ukiah is not the place for a 1200 foot link.  The
> fiber is not usually the problem.  It is the connectors on the ends of the
> fibers.  If you have a fusion splicer, it is not a problem.

I can recall watching the glass ends being meticulously polished for
individual cables.  Woe be unto he who cuts said cable "accidentally"!

> Oh, then there
> is the cost of the NICs.  The original plan was to get rid of a second
> satellite link costing about $100 a month.  Low data rates anyway.  They
> currently are using a long haul Wi-Fi connection down to Ukiah.  The
> 10Base-T connection has been in place about ten years now.

I have several fibre interfaces that I have set aside for a run
out to the back yard (looking for the extra isolation that they
afford)

> So far as "pushing-my-luck"  --  from Wikipedia:  (
> http://en.wikipedia.org/wiki/10BASE-T#Cabling)

Yeah, but 100 (or 150... or 200) meters of cable IN A BUILDING
is different from a geographical span of that same distance!

> Note there is no specific distance in the 10Base-T specification (150m is
> about 500ft).  The location, being will off the grid&  on the ocean side of
> the ridge, also provided a very low noise environment.

I'd be more worried about its exposure to lightning strikes
(not *direct* but, rather, *induced*) over such a long distance.
The galvanic isolation afforded by the coupling transformers
would be no match for an induced voltage spike from such a strike
(No idea how far into the machine any potential damage would
propagate).  I've lost (wired) phones from nearby strikes (the
surge takes out the protection diodes across the line) -- and
you *know* the phone company designs with this sort of thing
in mind!

But, I'm typically overly pessimistic about these things.  I
don't like *fixing* things if I can figure out a way to prevent
them from *breaking*, instead!