bridget ·
Ahead of test day on the 31st, I've been putting Traceport through some trial runs. Here's what Traceport is, what I tested, and what I found.
Traceport API is like a notary, but better. A notary stamps a document and says, "This person signed this on this day." But after that, the notary can't tell you whether anyone has changed the pages. With Traceport, anyone you share a record with can check for themselves that every word is exactly what was recorded. Change one comma and the check fails.
Here's how it works.
A company sends us a record, like a food shipment or a donation, and we put its digital fingerprint on the BSV blockchain. Later, an auditor, a regulator or a customer can check three things: that the record existed at that time, that it hasn't been changed, and which business it was recorded for. The proof is public, but the data isn't. The company keeps its records private and decides who gets to see and check them.
Trial Run.
Last weekend I tested whether it holds up at real volume. I put a hundred thousand records through it in one day, from many different businesses at once, including two of our real apps. 99,976 went through. Not one duplicate. Every successful record has its proof back from the blockchain.
The reason 100K is a good number for me: I didn't want to do just a blast demo. Firing a million transactions at the chain is easy if they all come from one key and you sign them ahead of time. I do think those tests are valuable. They show the network is fast and can handle big bursts. What they don't show is everything a real service has to do around each transaction. Traceport has to keep its own funds ready to pay for every record and top them up as it goes. It tracks every coin so nothing is spent twice. It signs each record with the right business's keys at the moment it arrives. And it collects the proof back from the blockchain, the SPV proof, so anyone with the record can verify it without trusting us. I wanted to test all of that.
How real customers would use it
So I ran it the way real customers would use it. I spread it over about 18 hours on purpose, at a steady pace with some busy spikes. Each business gets its own key space, cryptographically isolated from every other business, so one company's records can never be signed under another company's keys. Every successful record comes back with its proof.
About a sixth of one percent failed. I traced every one of them. Most came from one bug I found under load, a bookkeeping problem in how we tracked our own funds when the network was slow. No money was lost. I fixed it during the run, and the permanent fix is going in now. The rest were brief network blips. Nothing was double-recorded, and nothing went missing.
That's the point of a test like this: you find the thing before a customer does. On the 31st I'll run it again with the fixes in.