If there’s anything I can ever do to assist, I’m happy to do so, whether on a funding basis or learning to do some basic coding if that’s what’s needed here (I have no clue what this would entail, which will come as no surprise to you lol).
Thanks, man! I appreciate your willingness to help (seriously), that's very cool of you!!
It's not a funding problem (theymos pays me, and even if he didn't, there'd still be some subset of work, like this, that I'd be willing to do for free). It's also not a technical problem (I mean, it
is a technical problem, but, I'm confident that I either have already solved or can solve all of the issues related to this).
Basically, it's a
time problem: I've got a lot on my plate, and I'll likely only get to this late next year.
I didn't
plan on re-thinking how image uploading/hosting should really work on Bitcointalk, what happened was that I went through three evolutions of thought, initiated by the
Imgur issue:
(*) My first response to this (IIRC; it's been a while) was to re-host Loyce's Imgur backup on TalkImg, and then modify the image proxy to use some kind of lookup table to silently replace Imgur links with "equivalent" TalkImg links.
(*) My second thought (months later; again, IIRC) was to generalize my first thought and add a (pretty elaborate) hashing scheme to it so that there'd be no doubt as to the correctness of the "replaced" images. (I think I called this version of my thinking "CIPP" for "Community Image Preservation Project": I recall PMing theymos about it and getting a lukewarm-at-best response, so I shelved it.)
(*) My third (and current) thought is to leverage the very nice properties of
content addressability to solve the Imgur problem in a trustless way: I'd basically add a new BBCode tag (called
[cai], for "content-addressable image") that would be used to replace all of the Imgur
[img] tags with something that theymos could
independently construct (a text-encoded cryptographic hash of the image itself, so, something like:
[cai]26MrbcYgzHAo7kxxHSVxbtm6NMHn[/cai]). Getting the whole scheme to work properly involves the writing of two new pieces of software: a resolver-thingy that would be on the same network as the image proxy, and a storage-thingy that some members would (hopefully) choose to run on a volunteer-basis.
It just so happens that the way to comprehensively fix the Imgur problem
also puts most of the pieces in place for an image-hosting solution that would be basically perfect for Bitcointalk: one that
many users (in principle) can share the responsibility for, one that is tamper-proof by design (because you can't change images without also changing their hashes, which is what identifies them under this scheme in the first place), and one that beats the pants off something traditional (like
Chevereto) in terms of hosting costs and reliability (the storage nodes could live even without static IPs, much less anything DNS or SSL/TLS related, and the whole design is filled with little compute, storage, and bandwidth wins).
(I mean, there's an elephant in the room, and it's natural to think: "Erm, but, what about TalkImg?!". I'm not poo-pooing TalkImg, and I think JJ has done a great job with it, but, there are too many advantages to the approach I have in mind to ignore. Just like the Imgur images can all be transformed into
[cai] tags and absorbed in a seamless way, the TalkImg images could all be absorbed, too. Honestly, if I were JJ, I'd be kind of relieved that the long-term responsibility/upkeep and succession planning for TalkImg might just "dissolve".)