[c-nsp] SFP & GBIC module compatibility

[Phil Mayers]

On Thu, Dec 28, 2006 at 11:09:13PM -0500, Richard A Steenbergen wrote:
>On Thu, Dec 28, 2006 at 09:22:58PM +0000, Phil Mayers wrote:
>> >"Genuine Cisco" SFPs (that work) can be found online at about 70% or
>> >greater savings over Cisco on SX SFPs.  
>> 
>> Sorry, but IMNSHO, this is very bad advice. I and many others on this 
>> list have had extremely bad experiences with "cheap cisco" SFPs which 
>> are in fact forgeries.
>
>All the Cisco-labeled optics I've run across are counterfeit in name only 
>(i.e. they are produced on the same lines as the Cisco-labeled ones just 
>at night and under the table).

Some good points here. Comments below:

re: forgeries - lucky you. Your experience does not match mine. We had 
SFPs labelled as Cisco that variously:

 * would not autoneg with other equipment or themselves looped back

 * had visibly and measurably low TX light levels

 * had visibly distorted LC connectors, in particular the ceramic 
sleeves being offset (some were so distorted as to make inserting an LC 
patch lead physically impossible). In a great many cases, extremely high 
force had to be applied to unplug the LC patch lead

 * suffered very early life degradation manifesting as extremely high 
error rates on the order of a few percent

 * Had duplicate serial numbers - in one case, two SX and an LX SFP had 
the same serial number in the EEPROM. In no case was the serial on the 
label matched by the serial in the EEPROM.

>
>Personally I'd like to see people get off this "Genuine Cisco" kick. 

It was not my intention to suggest such, and I don't believe anyone else 
did?

"Genuine" vendor optics are an out-and-out scam, and not only would I 
urge everyone to purchase (high quality, for your sanity) generic optics 
(my recommendation: Prolabs), but I would urge you to do what we did at 
our last competitive tender and MANDATE support for such in your 
hardware.

We'll be having this discussion about Xenpak/X2/XFP in 2 years. Only the 
relatively low numbers (compared to gigE) make it a background issue.

>Nothing about optics having a Cisco label make them any more or less 
>genuine than any others. Do people really think that Cisco sends a team of 
>engineers to China to hand-pick only the finest SFP's fresh off the 
>production line, then polishes them to perfection and sprinkles them with 
>magic fairy dust before delivering them to your router on fluffy white 
>pillows?

I see what you're saying, but based on the differences between your and 
my experiences, I think you may be unaware of the fact there are 
certainly other sources of fake SFPs than just turning the crank on the 
production line 101 times but telling cisco you turned it 100 (and 
pocketing the difference).

>
>If anything, Cisco is responsible for the flood of substandard optics in 
>the marketplace. By implementing vendor locking, Cisco has prevented the 
>legitimate sale of competing optics directly to consumers. This makes 

I agree with that.

>
>It continues to amaze be how absurdly stupid and naive this industry is as 
>consumers. We are all the ones who are empowering Cisco (and everyone who 
>wants to be like Cisco) to abuse us as consumers, by buying in to their 

Speak for yourself! We make vendors work very hard through competitive 
tender at every major purchase (in fact, being a public sector institute 
we are mandated to do so, but after my experiences I would do it 
voluntarily). We've removed incumbents without rancour and declined 
Cisco offers on multiple occasions. We purchase the right tools for the 
job, and exclude proprietary protocols without a standards fallback in 
all our designs, to allow us that same choice.

I agree though that lots of people do think like this. I don't know how 
you'd solve it.


>fearmongering marketing tactics. Can you imagine an auto manufacturer who 

I've seen this analogy, and whilst it sort-of works it would be a lot 
more applicable if computers in general (and routers/switches are 
basically specialised computers) weren't total junk to start with. Cars 
are both a lot simpler and a lot more reliable than computers.

I could go on, but it's OT for this list.