Really, you can keep adding hashes to your log, but a client needs to save old hashes and have some access to old logs before they can verify anything.
IMO, a large reason everything went blockchain is you can tack it in without changing anything. Just setup a process to modify your logs and boom another buzzword.
Having global consensus (PoW) is a different matter, but that doesn't mean what you are saying isn't a blockchain.
There is a good conversation me and @geofft had the other day, would love your thoughts on it: https://news.ycombinator.com/item?id=17000401
In practice, many of these organizations talking about using in-house blockchains have legal trust relationships and systems of laws in common, and they won't only be relying on a transaction log to make sure the record isn't getting fiddled.
And often the threat model isn't related to changing the historical ledger anyhow; there is a lot of fraud, embezzlement, and other financial crime that attacks other points in the system. That requires a significant system of financial controls, which will be sufficient to catch ledger tampering as well. Meaning that the checksums become a modest nice-to-have, not a significant anti-fraud technique.
You are saying because these types of verified logs can't control for other variables, they do not represent a "complete" set of the system, and therefore aren't as useful for fraud analysis? Hmm.
But this is where I feel like calling a cryptocurrency economic model a "blockchain" is just a silly misappropriation of a word. To me, clearly "blockchain" is not like "capital" or "communes", a blockchain is a cryptographically verified data structure. However, I do hear a lot of people throwing around the word "blockchain" just to reference an economic model, which seems excessive.
I like your point though, and well described.
So, while Blockchain has nothing directly to do with being distributed, that's just a useful feature but requires proof of work of some kind, or someone can quickly manufacture several blocks.
Fair point about it being distributed
The Proof of Work solves a separate problem, that of Double-Spending. The Proof of Work here isn't arbitrarily to slow the system itself down, just to slow non-canonical histories.
But that itself is not the "blockchain" (thus the "how do you define it?" question, I'm curious to hear your take), I'm just playing devil's advocate: A lot of people who weren't around in the 70s are now calling verified logs a "blockchain", now it is a stupid buzzword, but at what point do we consolidate the word for an idea, versus argue there are semantic differences?