Hacker News new | past | comments | ask | show | jobs | submit login

Funny.

This is similar in concept to how Bitcoin defines its hardness, though it uses a different hash and block structure of course. I wonder if successful commits in this repo are coming from that world or maybe from SHA1 cryptographic weaknesses.




I have been bringing up this point in conversations about blockchain for a while now. Git is essentially a chain of commits, and the hash of a commit will change based on what you're applying it on by including the previous commit's hash in the body that gets eventually hashed into HEAD.

Also reminds me of that time back when we were using 8-char truncated git hashes as deployment tags and we eventually begun getting collisions. We also had an interesting bug when all 8 chars were numeric and Python (or the framework/library that was used) ingested what would be a string it as a number.


Git is a merkle tree. Blockchain also is a merkle tree. They are the same data structure.

https://en.wikipedia.org/wiki/Merkle_tree

The difference between blockhain and git is the hashing function. Git’s is designed for speed and constant difficulty, blockhain is designed for slowness and increasing difficulty. The logical meaning of each block is different also.


Python is mostly strongly typed. You can get a mix of floats and ints from e.g. parsing JSON, but your library's authors must have been pretty explicit on wanting to confuse strings with numbers.


I know, I was surprised too. But I guess if it comes in as a number from the get-go, it will simply continue as a number until we tried to slice it or something, which made the deployer abort. :)


I mean, of course the library in question could contain something like this:

    try:
        hash = int(raw_hash)
    except ValueError:
        hash = raw_hash
But WHY? Can you think of something one could do accidentally to achieve the same effect?


Maybe a migration from SVN and some legacy code somewhere. Since revisions will have r12345678 kind of format.


See, what it reminds me of is Knuth's bounty for errors.




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: