Larscom's CupreDSU product was an SDSL to V.35 DSU and has fascinated the interest of the Open SDSL Connectivity Project leader since December of 2004. It was only 5 years thereafter that we have finally been able to obtain a few of these units, and we now know that it is actually a rebranded XSB-2000 — see that page for more information.
What has later become the Open SDSL Connectivity Project started with a simple & innocent desire on my part to use a routed circuit configuration with V.35 hand-off for the SDSL service we were in the process of getting from what was then MCI. I knew next to nothing about SDSL at the time, but a search for compatible SDSL to V.35 CPE devices had revealed an extreme scarcity of such: there was Netopia D7100 which I had rejected as not being a true bit-transparent DSU, and there was Larscom CupreDSU.
So the next thing I wanted was to get a CupreDSU and live happily everafter with it. However, the article turned out to be pure unobtainium. I was told by Larscom (which still existed at the time) that only about 500 units had been made before the product was killed (they said no one was buying it) and I could not find any surplus equipment dealer that had such a unit.
Then I wanted to buy the schematics, BOM and gerber files for the CupreDSU
from Larscom, and although in my view even the most laughable amount was
still better than the absolute zero they were making from their unsellable
product, I couldn't offer them enough to be worth their time to deal with
the matter.
Once again I knew next to nothing about SDSL at the time, and I was hoping
that being able to peek at the CupreDSU schematics would answer my most basic
question of what the heck is this SDSL thing?
.
But not being able to buy those schematics, I was on my own.
Then after my interest in SDSL got put on hold and reopened some months later, I had learned about the Bt8970 and RS8973 bitpump chips being the foundation of SDSL/2B1Q. After forcibly liberating the source code for the bitpump's control software package and getting a little more insight into the variety of existing SDSL flavors by reverse-engineering a few of the existing CPE devices, I had realized that the CupreDSU and the Copper Mountain flavor it supported were just the tip of the iceberg.
That was when I had coined the name Open SDSL Connectivity Project and started out designing the OSDCU, which was then intended to serve as a debug instrument for cracking the secrets of non-understood SDSL flavors. But as the saying goes, if it wasn't hard, they wouldn't call it hardware, and the process of designing and building the OSDCU took quite a long time (3 years working on and off), an order of magnitude longer than it takes me to do a typical free software project. To make the long story short, I ended up doing the Hack-o-Rocket project and using it to accomplish the OSDCU's original goal (reverse-engineering unknown SDSL flavors) long before the OSDCU was finished.
But we did not abandon the OSDCU project, and in 2009 we had finally dug it out and finished building the board. Since its debug function (reverse-engineering SDSL flavors) had already been accomplished, the new focus was on the connectivity function, i.e., making an SDSL to EIA-530 (like V.35, but better) DSU. The OSDCU hardware is capable of performing exactly the same bit-transparent DSU function as the CupreDSU/XSB-2000, but is also envisioned as being able to serve as a Layer 2 converter from SDSL/ATM to HDLC.
And just when the OSDCU was finally fully built and we were getting ready to write the operational software that would make it work as an XSB-2000 equivalent, a bunch of XSB-2000 units (both CupreDSUs and the Lucent-branded equivalent) have landed in our lap!
A final comment needs to be made regarding the amount of money I had offered to Larscom back in 2004/2005. It's a reasonable assumption that they had laughed hysterically at my offer of $1000 for the manufacturing rights to their discontinued-due-to-unsellability CupreDSU, but consider this little factoid: we now have our own OSDCU hardware design which we have proven working and which we can manufacture in any desired quantity, and the total amount of money I have paid to other people to arrive at this design is less than $200.
The biggest cost was PCB layout: at first-world labor rates it would have cost somewhere around $2500, but a Russian comrade had offered to do it for 3000 Russian roubles (less than $100 USD) — he had told me it wasn't because he was so cheap, but because he supported Open Source Hardware! I've ended up paying him $150 (more than agreed upon initially) because he did such a great job. Adding Western Union transfer fees and other project expenses like me taking a coworker from my day job out to lunch as compensation for a design review, I still have spent less than $200 total to recreate on my own what I had offered to buy from Larscom for $1000 4.5 years prior.
Of course the figures above don't include the cost of parts, PCB fabrication or assembly, but none of that would have been included in the Larscom deal either — I was offering them $1000 for their gerber, schematic and BOM files only. The $200 figure also doesn't include my own labor on and off over the course of 3 years to create the OSDCU design prior to handing it over to layout, but that labor will never have any monetary amount associated with it as I am sworn to service to my Goddess and will continue to gift my labor to the Open Source and Free Computing community for as long as I breathe.
The $200 figure also doesn't include what I will soon have to pay to have a sheet metal enclosure designed and made for the OSDCU to turn it into a real product, but I wasn't counting on Larscom's plastics either, all I wanted from them was the bare board.