4 ms·

Ameo – Redis-compatible distributed key-value store with WebSocket API

40 points by networked 9y ago

chatmasta 9y ago
Cool. I implemented a similar toy project once, to directly expose a database on the server to the client via web socket. I say “toy” because that’s obviously incredibly insecure. What are the security considerations here? If you have multiple clients, should their topics be isolated? Presumably Alice should not be able to run a DEL command on a topic owned by Bob. Where would you put code for authentication / ACL?
- marianoguerra 9y ago
  the idea is to expose the websocket api to the public (for now only subscribe) and the redis api only to internal servers so they can publish/subscribe.
- goliatone 9y ago
  There are a couple projects that solve this by adding a rules engine that manages ACL. https://deepstreamhub.com https://deepstreamhub.com https://crossbar.io https://crossbar.io
- marknadal 9y ago
  We solved this by actually enabling P2P user accounts that secure the data and permissions through cryptographic signatures! https://hackernoon.com/so-you-want-to-build-a-p2p-twitter-with-e2e-encryption-f90505b2ff8 https://hackernoon.com/so-you-want-to-build-a-p2p-twitter-wi... This means the servers themselves can't even middleman the data.
- e12e 9y ago
  > I say “toy” because that’s obviously incredibly insecure. I trust postgres row-level access control a lot more than a php script with dbowner access to the db - so I think this depends a lot on which db, and what "direct access" means...
  - chatmasta 9y ago
    True. This was a college project years ago but recently I’ve been thinking about revisiting it. With properly provisioned DB users, ACLs, and stored procedures, it could actually end up being more secure than it would be with an intermediary application server.
Animats 9y ago
The title and article read like they were generated by a buzzword generator.
- marianoguerra 9y ago
  maybe, it's hard to write a meaningful title that describes the thing and it's still understandable. but all the technologies are used and for a good reason, no "throw technologies, mix it and see what comes out".
- dang 9y ago
  Submitted title was "Ameo – Redis-compatible distributed KV store on riak_core with WebSocket API"
edko 9y ago
Is ameo related in any way to altayanta?
- marianoguerra 9y ago
  linguistically yes, in any other way, I have no idea :)
stephen_mcd 9y ago
Cool! I built CurioDB a while back with the same idea, but using Akka as the underlying framework. Persistent Redis with distributed transactions, embedded Lua and a websocket front-end: https://github.com/stephenmcd/curiodb https://github.com/stephenmcd/curiodb
vardump 9y ago
Isn't WebSocket dead and buried, replaced by HTTP/2?
- netcraft 9y ago
  I hadnt heard this - after a bit of digging it looks like SSE + http/2 will replace a lot of the same use cases, but the consensus seems to be no: https://stackoverflow.com/questions/28582935/does-http-2-make-websockets-obsolete https://stackoverflow.com/questions/28582935/does-http-2-mak... Interesting thought though - I think more options there will be better in the long run, as long as it a "both" instead of an "either/or".

40 points by networked 9y ago

Cool. I implemented a similar toy project once, to directly expose a database on the server to the client via web socket. I say “toy” because that’s obviously incredibly insecure. What are the security considerations here? If you have multiple clients, should their topics be isolated? Presumably Alice should not be able to run a DEL command on a topic owned by Bob. Where would you put code for authentication / ACL?

marianoguerra 9y ago
the idea is to expose the websocket api to the public (for now only subscribe) and the redis api only to internal servers so they can publish/subscribe.
goliatone 9y ago
There are a couple projects that solve this by adding a rules engine that manages ACL. https://deepstreamhub.com https://deepstreamhub.com https://crossbar.io https://crossbar.io
marknadal 9y ago
We solved this by actually enabling P2P user accounts that secure the data and permissions through cryptographic signatures! https://hackernoon.com/so-you-want-to-build-a-p2p-twitter-with-e2e-encryption-f90505b2ff8 https://hackernoon.com/so-you-want-to-build-a-p2p-twitter-wi... This means the servers themselves can't even middleman the data.
e12e 9y ago
> I say “toy” because that’s obviously incredibly insecure. I trust postgres row-level access control a lot more than a php script with dbowner access to the db - so I think this depends a lot on which db, and what "direct access" means...
- chatmasta 9y ago
  True. This was a college project years ago but recently I’ve been thinking about revisiting it. With properly provisioned DB users, ACLs, and stored procedures, it could actually end up being more secure than it would be with an intermediary application server.

Animats 9y ago

The title and article read like they were generated by a buzzword generator.

marianoguerra 9y ago
maybe, it's hard to write a meaningful title that describes the thing and it's still understandable. but all the technologies are used and for a good reason, no "throw technologies, mix it and see what comes out".
dang 9y ago
Submitted title was "Ameo – Redis-compatible distributed KV store on riak_core with WebSocket API"

edko 9y ago

Is ameo related in any way to altayanta?

marianoguerra 9y ago
linguistically yes, in any other way, I have no idea :)

stephen_mcd 9y ago

Cool! I built CurioDB a while back with the same idea, but using Akka as the underlying framework. Persistent Redis with distributed transactions, embedded Lua and a websocket front-end: https://github.com/stephenmcd/curiodb https://github.com/stephenmcd/curiodb

vardump 9y ago

Isn't WebSocket dead and buried, replaced by HTTP/2?

netcraft 9y ago
I hadnt heard this - after a bit of digging it looks like SSE + http/2 will replace a lot of the same use cases, but the consensus seems to be no: https://stackoverflow.com/questions/28582935/does-http-2-make-websockets-obsolete https://stackoverflow.com/questions/28582935/does-http-2-mak... Interesting thought though - I think more options there will be better in the long run, as long as it a "both" instead of an "either/or".