Conversation
Notices
-
Self-hosting, free code, server app paradox: if devs set up a user-friendly demo instance, most people will use that instead of self-host
- Claes Wallin (韋嘉誠) and vinzv like this.
- Claes Wallin (韋嘉誠) repeated this.
-
@strypey Self-hosting needs to be as simple as running any other program---the user shouldn't have to host it on a "server". And ideally federated, so that the experience doesn't differ between using your instance and their own.
But yes, that is unfortunate indeed.
-
@mikegerwitz @strypey Federation usually means there is a server to federate with. True P2P distributed systems is where it's at.
-
@strypey Probably all demo servers need to self-destruct daily or weekly if you want to encourage self-hosting.
-
@strypey But then again "centralized federation" (the identi.ca syndrome) may have value when fighting monoliths. Best vs good, etc.
-
@clacke @mikegerwitz the problems with "true P2P" include power/ data costs for mobiles, PCs that aren't always on or online, and many more
-
@clacke that said, if as @mikegerwitz said a "server app" is an app on your PC like any other, it can support both P2P and federated uses
-
@clacke self-destructing demo servers make it hard for new projects to get the critical mass of users to benefit from a network effect
-
@strypey @mikegerwitz Yes. If it was easy, P2P would have boomed a long time ago. Another biggie: search engine indexability.
-
@strypey @mikegerwitz Yes. A successful distributed service probably needs to combine the P2P and federated/centralized aspects.
-
@strypey Yes.
-
@clacke @mikegerwitz good point, though this is more important for public-facing publishing and events than for private comms, file sync etc
-
@strypey @mikegerwitz For social networks I believe it's a biggie. When people have public conversations, they add value to the Web.
-
@strypey @mikegerwitz When people find discussions, it promotes the network. GNUSocial sometimes provides search hits for me. pump.io never.
-
@clacke @mikegerwitz I keep thinking about federated search. GS has a search function, as do most web apps. Could a server meta-search them?
-
Sorry, Lambda is a French term to design generic/normal people without experience in X domain, I thought it meant the same.
-
@mangeurdenuage @strypey So, something like lay people?
-
@strypey @clacke I hadn't given too much thought to search, but it seems like indexing federated systems wouldn't be something too difficult to implement, provided that an indexer deduplicates records and points the user to the authoritative source of the record. If I search for something in i.e. Google, that's one of the major issues: same text on so many separate instances.
P2P/Mesh is more interesting, but could be reduced to a federated search problem if enough data propagates to nodes persistent enough to be discovered by other indexers (e.g. Google). But then the user might be directed to a node that's not online, so what would the fallback be? A cache? Would that cache be centralized? Federated? What should be trusted with such authority? If it's cryptographically signed by the originating node, maybe it won't matter, provided a proper trust model.
If anyone has experience thinking about these things, I'd be curious to know; I haven't done research myself.
-
@clacke good thoughts @mikegerwitz I know 0 about how to implement this but it just seems like letting search remain centralized is unwise
-
@strypey @clacke @mikegerwitz it has sometimes been said that web search is a fundamentally centralized thing, but given DHTs and such I'm doubtful about that.
-
@bob @strypey @mikegerwitz Trusting results is the only issue I see that "requires" centralization. Redundant verification should fix that.
-
@mikegerwitz @strypey A central or distributed firehose with some publically-knowable web-readable nodes should be able to index a P2P net.
-
something like the blockchain servers that software like electrum use
-
@mikael Yes, consensus protocols come to mind.
-
@clacke @mikegerwitz could something like #yacy or the various orphaned P2P search projects fit into this vision?
http://qttr.at/1h2o
-
@strypey @mikegerwitz Meta-searching application-specific engines is an interesting idea. Won't help general public visibility though.
-
@clacke @mikegerwitz I'm imagining something like #yacy using meta-search to populate its index.
-
@clacke @mikegerwitz ...or Searx, a meta-search engine licensed under AGPL:
http://qttr.at/1h45
-
@strypey Thanks for sharing your research! I think a P2P search is essential, but we also need a way for it to feed into the indexes of the larger centralized search engines that everyone else uses. I hope one day that won't be necessary, but until that time, others need to be able to use it to discover. :(
A standard protocol is also important; free/libre projects are great, but fragmentation will stall this into oblivion.
But I'm optimistic!
cc @clacke
-
@mikegerwitz @clacke agree 100% on open standards but its a chicken/egg problem...
-
@clacke @mikegerwitz support for standards emerges from consensus among devs, who often want to hack on their app, not negotiate standards
-
@clacke @mikegerwitz that's what it looks like from the POV of a novice coder whose contribution to free code so far is evangelism not dev
-
@strypey @mikegerwitz There's a delicate balance involved, between innovation and crystallization. Projects must benefit from standards.