News for hackers. Stuff that matters.
Re:Distributing NixOS with IPFS by rkeene2 on Thursday January 19, 2017 @11:47AM (#13437195)
Good to see other people are inventing AppFS ( http://appfs.rkeene.org/ ) :-)
Re:Distributing NixOS with IPFS by _prometheus on Friday January 20, 2017 @09:10PM (#13448387)
> Good to see other people are inventing AppFS ( http://appfs.rkeene.org/ ) :-)

I'll take the construction "other people are inventing <the thing i invented some time ago>" to mean that you think you came up with this first, or at least prior to Nix or IPFS. And thus ":-)" to be a bit sarcastic and unhappy, instead of genuinely happy.

Similar times:

* http://appfs.rkeene.org/web/timeline?c=78c60b0c9e7da1c9&unhi...

* https://gist.github.com/jbenet/8f000606f2009495c56177f6ca2c1...

* https://github.com/ipfs/ipfs/commit/8004db75262fcd29399d6c7f...

* https://github.com/jbenet/random-ideas/issues/19

* i think the Nix people probably came up with this way before either of us

* and i am willing to bet everything that at least 100 other people (maybe thousands) have came up with this exact same idea over a decade before any of us...

In fact, most of the best "original ideas" in the IPFS body of work were probably first discovered decades prior.

I repeatedly see people succumbing to sadness over multiple discovery. It shouldn't be sad, it should be a happy event, as it confirms our own thoughts and presents an opportunity for collaborations. :)

Further thoughts here: https://gist.github.com/jbenet/8f000606f2009495c56177f6ca2c1...

Re:Distributing NixOS with IPFS by rkeene2 on Tuesday January 31, 2017 @07:50PM (#13537061)
1. I was genuinely happy that people are doing this. I'm happy because it's good for me (I can run more software in the way I prefer to, without installing it). I never expressed anything emotional icon other than the "smiley face", which does not denote sadness.

2. The items with similar timelines are not comparable. The link to AppFS is a link to working code being written and the other links are to documents regarding a desire to have working code written someday, maybe.

3. It's a good idea, that I did not claim to come up with. It's a refinement of how things were done on UNIX campuses long ago where the was a common NFS-mounted directory with applications installed on it. This is definitely not original out claimed to be do. Then the same model that Sun/Oracle uses for Solaris packages in "ipkg" in Solaris 11 is mixed in with that. The result is something very similar to something old as well called 0install, back when it used LazyFS. Again not original. The thing that AppFS adds beyond just the merging of these ideas is the additional idea that the filesystem should be writable. This is not original exactly either since the same thing is done for ClusterNFS (except per-system changes are preserved instead of per-uid).

Given that I've obviously used all these technologies to achieve similar goals it should be obvious that I don't think AppFS is original -- quite the opposite, I've been using it for decades.

The intent of my post, which you seem to have missed, is to point to something else that does the same thing as this new thing. The reason that I would do this is because the ideas a are old and we can improve upon them by looking at similar implementations. I certainly spent a lot of time with 0install's LazyFS (probably one of the heaviest packages).

As an aside, I found your post condescending in tone. Given that you acknowledged that all your statements were based solely on assumptions for things I did not write, I do not think this was effective in communicating your thoughts over this medium. Overall, I'm not sure what your message would contribute if these assumptions you made were wrong. My guess (assumption) is that you did not consider what your message was if your assumptions were wrong. If my assumption/guess is wrong here, please disregard this section and do note that this message still has content that communicates meaning.

Re:Distributing NixOS with IPFS by HurrdurrHodor on Thursday January 19, 2017 @11:58AM (#13437275)
Please don't use SHA-1. It's almost broken.
Re:Distributing NixOS with IPFS by rkeene2 on Thursday January 19, 2017 @12:36PM (#13437571)
The protocol is actually extensible, and the hashing algorithm MUST always be specified by the server (which the client could then choose to not accept, just as it can reject the certificate because of the signature algorithm).

Also, it would require a preimage attack against one of the hashed items to be useful which SHA-1 will likely be resistant to a long time (though decreasing with the number of items hashed) and SHA-1 is unlikely to be vulnerable to a preimage attack in the near future based on what we know so far.

The signature and certificates that are used to validate the top-level index can be based on a far better hashing algorithm independently of the content-based hashing.

Re:Distributing NixOS with IPFS by lgierth on Thursday January 19, 2017 @12:26PM (#13437488)
Use multihashes for hash algorithm agility :) https://github.com/multiformats/multihash
Re:Distributing NixOS with IPFS by rkeene2 on Thursday January 19, 2017 @12:38PM (#13437595)
Multiple hashing algorithms is already built-in and mandatory, everywhere a hashing operation is used the hashing algorithm must also be specified.